diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,84021 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.1653903778924588,
+  "eval_steps": 500,
+  "global_step": 14000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.242144177449168e-08,
+      "loss": 17.0871,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8484288354898337e-07,
+      "loss": 16.7584,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7726432532347505e-07,
+      "loss": 18.7884,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6968576709796674e-07,
+      "loss": 16.8019,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.621072088724585e-07,
+      "loss": 17.7753,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.545286506469501e-07,
+      "loss": 18.6416,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.469500924214418e-07,
+      "loss": 16.358,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.393715341959335e-07,
+      "loss": 16.7555,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.317929759704252e-07,
+      "loss": 17.0,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.24214417744917e-07,
+      "loss": 17.2552,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0166358595194085e-06,
+      "loss": 16.1383,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1090573012939002e-06,
+      "loss": 17.6004,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.201478743068392e-06,
+      "loss": 16.6063,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2939001848428836e-06,
+      "loss": 17.329,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3863216266173753e-06,
+      "loss": 17.2745,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.478743068391867e-06,
+      "loss": 15.1228,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5711645101663589e-06,
+      "loss": 18.54,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6635859519408503e-06,
+      "loss": 19.1136,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.756007393715342e-06,
+      "loss": 16.5361,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.848428835489834e-06,
+      "loss": 17.4019,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9408502772643254e-06,
+      "loss": 16.6949,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.033271719038817e-06,
+      "loss": 15.8961,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1256931608133088e-06,
+      "loss": 15.7922,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2181146025878004e-06,
+      "loss": 16.7225,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.310536044362292e-06,
+      "loss": 17.8774,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.402957486136784e-06,
+      "loss": 16.8348,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4953789279112755e-06,
+      "loss": 14.7975,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.587800369685767e-06,
+      "loss": 20.769,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.680221811460259e-06,
+      "loss": 18.233,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7726432532347505e-06,
+      "loss": 19.9336,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8650646950092422e-06,
+      "loss": 19.2667,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.957486136783734e-06,
+      "loss": 16.808,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0499075785582256e-06,
+      "loss": 15.7567,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1423290203327177e-06,
+      "loss": 19.1186,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.234750462107209e-06,
+      "loss": 17.565,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.3271719038817007e-06,
+      "loss": 18.0256,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.4195933456561928e-06,
+      "loss": 17.6757,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.512014787430684e-06,
+      "loss": 15.8883,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6044362292051757e-06,
+      "loss": 17.5371,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.696857670979668e-06,
+      "loss": 17.8567,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.789279112754159e-06,
+      "loss": 15.5156,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.881700554528651e-06,
+      "loss": 16.0591,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.9741219963031425e-06,
+      "loss": 16.0072,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.066543438077634e-06,
+      "loss": 20.3638,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.158964879852126e-06,
+      "loss": 19.0529,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.2513863216266175e-06,
+      "loss": 16.772,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.343807763401109e-06,
+      "loss": 20.3749,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.436229205175601e-06,
+      "loss": 15.7362,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.5286506469500926e-06,
+      "loss": 15.4111,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.621072088724584e-06,
+      "loss": 18.4005,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.713493530499076e-06,
+      "loss": 17.6586,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.805914972273568e-06,
+      "loss": 16.0131,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.898336414048059e-06,
+      "loss": 18.2993,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.990757855822551e-06,
+      "loss": 18.3884,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.083179297597043e-06,
+      "loss": 17.0226,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.175600739371534e-06,
+      "loss": 16.7854,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.268022181146026e-06,
+      "loss": 17.2643,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.360443622920518e-06,
+      "loss": 16.3639,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.452865064695009e-06,
+      "loss": 18.845,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.545286506469501e-06,
+      "loss": 17.6615,
+      "step": 60
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.637707948243993e-06,
+      "loss": 18.5154,
+      "step": 61
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.7301293900184845e-06,
+      "loss": 17.9647,
+      "step": 62
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.822550831792976e-06,
+      "loss": 20.0969,
+      "step": 63
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.914972273567468e-06,
+      "loss": 18.17,
+      "step": 64
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.0073937153419595e-06,
+      "loss": 14.6499,
+      "step": 65
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.099815157116451e-06,
+      "loss": 18.0505,
+      "step": 66
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.192236598890943e-06,
+      "loss": 17.6813,
+      "step": 67
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.284658040665435e-06,
+      "loss": 18.5939,
+      "step": 68
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.377079482439926e-06,
+      "loss": 17.036,
+      "step": 69
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.469500924214418e-06,
+      "loss": 16.7875,
+      "step": 70
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.56192236598891e-06,
+      "loss": 18.4579,
+      "step": 71
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.654343807763401e-06,
+      "loss": 18.082,
+      "step": 72
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.746765249537892e-06,
+      "loss": 17.4138,
+      "step": 73
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.8391866913123855e-06,
+      "loss": 19.6673,
+      "step": 74
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.931608133086876e-06,
+      "loss": 18.7293,
+      "step": 75
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.024029574861368e-06,
+      "loss": 19.1737,
+      "step": 76
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.11645101663586e-06,
+      "loss": 16.7308,
+      "step": 77
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.2088724584103514e-06,
+      "loss": 19.8483,
+      "step": 78
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.301293900184842e-06,
+      "loss": 16.5522,
+      "step": 79
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.393715341959336e-06,
+      "loss": 17.9661,
+      "step": 80
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.486136783733827e-06,
+      "loss": 18.889,
+      "step": 81
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.578558225508318e-06,
+      "loss": 17.5664,
+      "step": 82
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.67097966728281e-06,
+      "loss": 18.456,
+      "step": 83
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.763401109057302e-06,
+      "loss": 18.438,
+      "step": 84
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.855822550831792e-06,
+      "loss": 17.4639,
+      "step": 85
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.948243992606285e-06,
+      "loss": 19.0733,
+      "step": 86
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.040665434380777e-06,
+      "loss": 16.9684,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.133086876155268e-06,
+      "loss": 17.767,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.22550831792976e-06,
+      "loss": 19.3689,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.317929759704252e-06,
+      "loss": 16.7624,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.410351201478742e-06,
+      "loss": 17.7719,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.502772643253235e-06,
+      "loss": 17.8388,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.595194085027728e-06,
+      "loss": 17.9006,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.687615526802218e-06,
+      "loss": 17.6653,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.780036968576711e-06,
+      "loss": 19.0401,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.872458410351202e-06,
+      "loss": 19.1006,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.964879852125693e-06,
+      "loss": 17.9726,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.057301293900185e-06,
+      "loss": 17.6885,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.149722735674678e-06,
+      "loss": 16.6701,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.242144177449168e-06,
+      "loss": 17.5058,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.334565619223661e-06,
+      "loss": 15.9845,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.426987060998152e-06,
+      "loss": 16.45,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.519408502772643e-06,
+      "loss": 16.6629,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.611829944547135e-06,
+      "loss": 17.3197,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.704251386321626e-06,
+      "loss": 19.3779,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.796672828096119e-06,
+      "loss": 17.323,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.889094269870611e-06,
+      "loss": 15.5748,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.981515711645102e-06,
+      "loss": 17.1617,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0073937153419593e-05,
+      "loss": 17.5135,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0166358595194085e-05,
+      "loss": 16.7948,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0258780036968576e-05,
+      "loss": 18.6314,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0351201478743069e-05,
+      "loss": 18.3545,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0443622920517561e-05,
+      "loss": 15.6468,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0536044362292052e-05,
+      "loss": 17.2081,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0628465804066545e-05,
+      "loss": 18.1747,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0720887245841035e-05,
+      "loss": 16.7779,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0813308687615526e-05,
+      "loss": 17.3272,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0905730129390019e-05,
+      "loss": 18.1059,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0998151571164511e-05,
+      "loss": 16.8923,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1090573012939002e-05,
+      "loss": 18.1103,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1182994454713495e-05,
+      "loss": 15.2079,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1275415896487986e-05,
+      "loss": 19.9749,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1367837338262476e-05,
+      "loss": 17.8166,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1460258780036969e-05,
+      "loss": 17.6362,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1552680221811461e-05,
+      "loss": 14.2711,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1645101663585952e-05,
+      "loss": 16.3334,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1737523105360445e-05,
+      "loss": 20.6417,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1829944547134936e-05,
+      "loss": 17.1523,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1922365988909427e-05,
+      "loss": 18.0802,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2014787430683919e-05,
+      "loss": 16.8187,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2107208872458412e-05,
+      "loss": 16.3878,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2199630314232902e-05,
+      "loss": 18.1122,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2292051756007395e-05,
+      "loss": 16.7671,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2384473197781886e-05,
+      "loss": 16.2677,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2476894639556377e-05,
+      "loss": 16.2427,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.256931608133087e-05,
+      "loss": 18.1572,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.266173752310536e-05,
+      "loss": 17.1701,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2754158964879853e-05,
+      "loss": 18.604,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2846580406654343e-05,
+      "loss": 16.5119,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2939001848428836e-05,
+      "loss": 17.1828,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3031423290203328e-05,
+      "loss": 18.0694,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.312384473197782e-05,
+      "loss": 16.5675,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3216266173752312e-05,
+      "loss": 18.0868,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3308687615526803e-05,
+      "loss": 16.9491,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3401109057301295e-05,
+      "loss": 16.4859,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3493530499075784e-05,
+      "loss": 17.4982,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3585951940850279e-05,
+      "loss": 17.515,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3678373382624771e-05,
+      "loss": 16.6041,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.377079482439926e-05,
+      "loss": 17.4948,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3863216266173753e-05,
+      "loss": 18.3241,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3955637707948244e-05,
+      "loss": 17.1114,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4048059149722736e-05,
+      "loss": 17.0734,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4140480591497229e-05,
+      "loss": 18.7625,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.423290203327172e-05,
+      "loss": 17.679,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4325323475046212e-05,
+      "loss": 19.4289,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4417744916820703e-05,
+      "loss": 17.4586,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4510166358595195e-05,
+      "loss": 17.3837,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4602587800369685e-05,
+      "loss": 17.3428,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4695009242144179e-05,
+      "loss": 16.1485,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4787430683918671e-05,
+      "loss": 16.868,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.487985212569316e-05,
+      "loss": 17.1999,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4972273567467655e-05,
+      "loss": 19.6471,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5064695009242144e-05,
+      "loss": 17.3717,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5157116451016636e-05,
+      "loss": 18.0381,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5249537892791127e-05,
+      "loss": 18.161,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.534195933456562e-05,
+      "loss": 17.4885,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5434380776340114e-05,
+      "loss": 18.0149,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5526802218114603e-05,
+      "loss": 21.9571,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5619223659889096e-05,
+      "loss": 17.3684,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5711645101663585e-05,
+      "loss": 17.978,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5804066543438077e-05,
+      "loss": 20.0705,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.589648798521257e-05,
+      "loss": 18.5101,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5988909426987062e-05,
+      "loss": 17.7676,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6081330868761555e-05,
+      "loss": 17.8342,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6173752310536044e-05,
+      "loss": 15.8783,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6266173752310537e-05,
+      "loss": 17.3738,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.635859519408503e-05,
+      "loss": 18.379,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.645101663585952e-05,
+      "loss": 18.6527,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6543438077634014e-05,
+      "loss": 18.5231,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6635859519408503e-05,
+      "loss": 18.6162,
+      "step": 180
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6728280961182996e-05,
+      "loss": 18.0102,
+      "step": 181
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6820702402957485e-05,
+      "loss": 18.0115,
+      "step": 182
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6913123844731977e-05,
+      "loss": 18.6351,
+      "step": 183
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.700554528650647e-05,
+      "loss": 18.3997,
+      "step": 184
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7097966728280963e-05,
+      "loss": 18.7155,
+      "step": 185
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7190388170055455e-05,
+      "loss": 17.4315,
+      "step": 186
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7282809611829944e-05,
+      "loss": 16.5158,
+      "step": 187
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7375231053604437e-05,
+      "loss": 16.1004,
+      "step": 188
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.746765249537893e-05,
+      "loss": 18.4571,
+      "step": 189
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7560073937153422e-05,
+      "loss": 17.5456,
+      "step": 190
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7652495378927914e-05,
+      "loss": 16.6723,
+      "step": 191
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7744916820702404e-05,
+      "loss": 18.0336,
+      "step": 192
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7837338262476896e-05,
+      "loss": 17.7359,
+      "step": 193
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7929759704251385e-05,
+      "loss": 16.625,
+      "step": 194
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8022181146025878e-05,
+      "loss": 17.4342,
+      "step": 195
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.811460258780037e-05,
+      "loss": 18.6018,
+      "step": 196
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8207024029574863e-05,
+      "loss": 17.8335,
+      "step": 197
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8299445471349355e-05,
+      "loss": 16.6765,
+      "step": 198
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8391866913123844e-05,
+      "loss": 16.5366,
+      "step": 199
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8484288354898337e-05,
+      "loss": 17.3106,
+      "step": 200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.857670979667283e-05,
+      "loss": 17.4574,
+      "step": 201
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8669131238447322e-05,
+      "loss": 18.5809,
+      "step": 202
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.876155268022181e-05,
+      "loss": 17.2708,
+      "step": 203
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8853974121996304e-05,
+      "loss": 18.9843,
+      "step": 204
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8946395563770796e-05,
+      "loss": 17.8682,
+      "step": 205
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9038817005545285e-05,
+      "loss": 21.4451,
+      "step": 206
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.913123844731978e-05,
+      "loss": 16.6807,
+      "step": 207
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.922365988909427e-05,
+      "loss": 16.5174,
+      "step": 208
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9316081330868763e-05,
+      "loss": 17.9328,
+      "step": 209
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9408502772643252e-05,
+      "loss": 16.1074,
+      "step": 210
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9500924214417745e-05,
+      "loss": 19.6828,
+      "step": 211
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9593345656192237e-05,
+      "loss": 17.3814,
+      "step": 212
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.968576709796673e-05,
+      "loss": 16.2921,
+      "step": 213
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9778188539741222e-05,
+      "loss": 17.222,
+      "step": 214
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987060998151571e-05,
+      "loss": 17.574,
+      "step": 215
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9963031423290204e-05,
+      "loss": 18.0833,
+      "step": 216
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.0055452865064696e-05,
+      "loss": 16.3459,
+      "step": 217
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.0147874306839186e-05,
+      "loss": 16.4422,
+      "step": 218
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.024029574861368e-05,
+      "loss": 17.4017,
+      "step": 219
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.033271719038817e-05,
+      "loss": 17.4014,
+      "step": 220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.0425138632162663e-05,
+      "loss": 15.5418,
+      "step": 221
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.0517560073937152e-05,
+      "loss": 15.6842,
+      "step": 222
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.0609981515711645e-05,
+      "loss": 18.2224,
+      "step": 223
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.0702402957486137e-05,
+      "loss": 15.6112,
+      "step": 224
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.079482439926063e-05,
+      "loss": 17.1537,
+      "step": 225
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.0887245841035122e-05,
+      "loss": 18.6784,
+      "step": 226
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.097966728280961e-05,
+      "loss": 18.2757,
+      "step": 227
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.1072088724584104e-05,
+      "loss": 19.4959,
+      "step": 228
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.1164510166358593e-05,
+      "loss": 19.7008,
+      "step": 229
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.125693160813309e-05,
+      "loss": 16.0107,
+      "step": 230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.1349353049907582e-05,
+      "loss": 17.6328,
+      "step": 231
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.144177449168207e-05,
+      "loss": 18.4515,
+      "step": 232
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.1534195933456563e-05,
+      "loss": 17.6586,
+      "step": 233
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.1626617375231053e-05,
+      "loss": 16.5373,
+      "step": 234
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.1719038817005545e-05,
+      "loss": 17.2454,
+      "step": 235
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.1811460258780038e-05,
+      "loss": 18.6714,
+      "step": 236
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.190388170055453e-05,
+      "loss": 18.5115,
+      "step": 237
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.1996303142329023e-05,
+      "loss": 18.0527,
+      "step": 238
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2088724584103512e-05,
+      "loss": 17.835,
+      "step": 239
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2181146025878004e-05,
+      "loss": 18.1486,
+      "step": 240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2273567467652497e-05,
+      "loss": 19.852,
+      "step": 241
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.236598890942699e-05,
+      "loss": 19.3144,
+      "step": 242
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2458410351201482e-05,
+      "loss": 18.1993,
+      "step": 243
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.255083179297597e-05,
+      "loss": 17.8801,
+      "step": 244
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2643253234750464e-05,
+      "loss": 18.9704,
+      "step": 245
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2735674676524953e-05,
+      "loss": 17.3948,
+      "step": 246
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2828096118299445e-05,
+      "loss": 18.4636,
+      "step": 247
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2920517560073938e-05,
+      "loss": 17.0178,
+      "step": 248
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.301293900184843e-05,
+      "loss": 18.3165,
+      "step": 249
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3105360443622923e-05,
+      "loss": 19.8496,
+      "step": 250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3197781885397412e-05,
+      "loss": 17.5555,
+      "step": 251
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3290203327171905e-05,
+      "loss": 19.149,
+      "step": 252
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3382624768946397e-05,
+      "loss": 17.5899,
+      "step": 253
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.347504621072089e-05,
+      "loss": 18.3817,
+      "step": 254
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3567467652495382e-05,
+      "loss": 15.544,
+      "step": 255
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.365988909426987e-05,
+      "loss": 17.6606,
+      "step": 256
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3752310536044364e-05,
+      "loss": 17.1198,
+      "step": 257
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3844731977818853e-05,
+      "loss": 17.2499,
+      "step": 258
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3937153419593346e-05,
+      "loss": 16.985,
+      "step": 259
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.4029574861367838e-05,
+      "loss": 18.0131,
+      "step": 260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.412199630314233e-05,
+      "loss": 16.4685,
+      "step": 261
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.4214417744916823e-05,
+      "loss": 17.1998,
+      "step": 262
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.4306839186691312e-05,
+      "loss": 16.7435,
+      "step": 263
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.4399260628465805e-05,
+      "loss": 18.205,
+      "step": 264
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.4491682070240297e-05,
+      "loss": 17.9308,
+      "step": 265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.458410351201479e-05,
+      "loss": 18.8333,
+      "step": 266
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.467652495378928e-05,
+      "loss": 17.7437,
+      "step": 267
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.476894639556377e-05,
+      "loss": 18.234,
+      "step": 268
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.4861367837338264e-05,
+      "loss": 18.502,
+      "step": 269
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.4953789279112753e-05,
+      "loss": 17.8266,
+      "step": 270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.504621072088725e-05,
+      "loss": 17.0264,
+      "step": 271
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.513863216266174e-05,
+      "loss": 16.7708,
+      "step": 272
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5231053604436227e-05,
+      "loss": 17.1925,
+      "step": 273
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.532347504621072e-05,
+      "loss": 16.7207,
+      "step": 274
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5415896487985213e-05,
+      "loss": 19.6391,
+      "step": 275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5508317929759705e-05,
+      "loss": 18.2234,
+      "step": 276
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5600739371534198e-05,
+      "loss": 17.0619,
+      "step": 277
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5693160813308687e-05,
+      "loss": 16.1002,
+      "step": 278
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.578558225508318e-05,
+      "loss": 18.4552,
+      "step": 279
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5878003696857672e-05,
+      "loss": 18.1973,
+      "step": 280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5970425138632164e-05,
+      "loss": 18.3274,
+      "step": 281
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.6062846580406657e-05,
+      "loss": 19.1444,
+      "step": 282
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.6155268022181146e-05,
+      "loss": 18.3889,
+      "step": 283
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.624768946395564e-05,
+      "loss": 20.1959,
+      "step": 284
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.634011090573013e-05,
+      "loss": 17.0318,
+      "step": 285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.6432532347504624e-05,
+      "loss": 16.9369,
+      "step": 286
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.6524953789279116e-05,
+      "loss": 18.9097,
+      "step": 287
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.6617375231053605e-05,
+      "loss": 18.0962,
+      "step": 288
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.6709796672828098e-05,
+      "loss": 18.7542,
+      "step": 289
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.680221811460259e-05,
+      "loss": 19.8932,
+      "step": 290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.6894639556377083e-05,
+      "loss": 16.694,
+      "step": 291
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.698706099815157e-05,
+      "loss": 18.2648,
+      "step": 292
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.707948243992606e-05,
+      "loss": 18.811,
+      "step": 293
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7171903881700557e-05,
+      "loss": 18.7483,
+      "step": 294
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.726432532347505e-05,
+      "loss": 17.67,
+      "step": 295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7356746765249542e-05,
+      "loss": 17.9565,
+      "step": 296
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7449168207024028e-05,
+      "loss": 18.9209,
+      "step": 297
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.754158964879852e-05,
+      "loss": 18.2572,
+      "step": 298
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7634011090573013e-05,
+      "loss": 18.974,
+      "step": 299
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7726432532347505e-05,
+      "loss": 18.0486,
+      "step": 300
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.7818853974122e-05,
+      "loss": 17.8596,
+      "step": 301
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.7911275415896487e-05,
+      "loss": 16.5102,
+      "step": 302
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.800369685767098e-05,
+      "loss": 18.5856,
+      "step": 303
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.8096118299445472e-05,
+      "loss": 17.3765,
+      "step": 304
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.8188539741219965e-05,
+      "loss": 16.8757,
+      "step": 305
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.8280961182994457e-05,
+      "loss": 18.5371,
+      "step": 306
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.8373382624768946e-05,
+      "loss": 17.7881,
+      "step": 307
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.846580406654344e-05,
+      "loss": 20.3273,
+      "step": 308
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.855822550831793e-05,
+      "loss": 16.9988,
+      "step": 309
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.8650646950092424e-05,
+      "loss": 17.869,
+      "step": 310
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.8743068391866913e-05,
+      "loss": 17.6561,
+      "step": 311
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.8835489833641406e-05,
+      "loss": 17.9161,
+      "step": 312
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.8927911275415898e-05,
+      "loss": 17.771,
+      "step": 313
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.902033271719039e-05,
+      "loss": 17.1554,
+      "step": 314
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9112754158964883e-05,
+      "loss": 16.38,
+      "step": 315
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.920517560073937e-05,
+      "loss": 17.5334,
+      "step": 316
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9297597042513865e-05,
+      "loss": 16.7254,
+      "step": 317
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9390018484288358e-05,
+      "loss": 19.4825,
+      "step": 318
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.948243992606285e-05,
+      "loss": 16.7767,
+      "step": 319
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9574861367837343e-05,
+      "loss": 18.9677,
+      "step": 320
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.966728280961183e-05,
+      "loss": 14.5711,
+      "step": 321
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.975970425138632e-05,
+      "loss": 17.2589,
+      "step": 322
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9852125693160817e-05,
+      "loss": 16.806,
+      "step": 323
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.994454713493531e-05,
+      "loss": 18.1543,
+      "step": 324
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.0036968576709802e-05,
+      "loss": 17.9545,
+      "step": 325
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.0129390018484288e-05,
+      "loss": 16.4624,
+      "step": 326
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.022181146025878e-05,
+      "loss": 16.8447,
+      "step": 327
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.0314232902033273e-05,
+      "loss": 15.488,
+      "step": 328
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.0406654343807765e-05,
+      "loss": 17.8605,
+      "step": 329
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.0499075785582254e-05,
+      "loss": 16.9954,
+      "step": 330
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.059149722735675e-05,
+      "loss": 17.8687,
+      "step": 331
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.068391866913124e-05,
+      "loss": 16.9587,
+      "step": 332
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.077634011090573e-05,
+      "loss": 16.7957,
+      "step": 333
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.086876155268023e-05,
+      "loss": 16.1118,
+      "step": 334
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.096118299445471e-05,
+      "loss": 17.9746,
+      "step": 335
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.1053604436229206e-05,
+      "loss": 17.8442,
+      "step": 336
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.1146025878003695e-05,
+      "loss": 16.3142,
+      "step": 337
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.123844731977819e-05,
+      "loss": 18.007,
+      "step": 338
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.133086876155269e-05,
+      "loss": 17.041,
+      "step": 339
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.142329020332717e-05,
+      "loss": 16.5292,
+      "step": 340
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.1515711645101665e-05,
+      "loss": 17.9467,
+      "step": 341
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.1608133086876155e-05,
+      "loss": 17.1737,
+      "step": 342
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.170055452865065e-05,
+      "loss": 15.992,
+      "step": 343
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.179297597042514e-05,
+      "loss": 17.8831,
+      "step": 344
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.188539741219963e-05,
+      "loss": 17.3478,
+      "step": 345
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.1977818853974125e-05,
+      "loss": 16.367,
+      "step": 346
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.2070240295748614e-05,
+      "loss": 16.7848,
+      "step": 347
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.216266173752311e-05,
+      "loss": 18.3435,
+      "step": 348
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.22550831792976e-05,
+      "loss": 16.1696,
+      "step": 349
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.234750462107209e-05,
+      "loss": 17.7612,
+      "step": 350
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.2439926062846584e-05,
+      "loss": 16.2172,
+      "step": 351
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.253234750462107e-05,
+      "loss": 18.3958,
+      "step": 352
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.262476894639557e-05,
+      "loss": 18.7539,
+      "step": 353
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.271719038817006e-05,
+      "loss": 17.7665,
+      "step": 354
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.280961182994455e-05,
+      "loss": 18.5671,
+      "step": 355
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.290203327171904e-05,
+      "loss": 18.3255,
+      "step": 356
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.299445471349353e-05,
+      "loss": 18.5766,
+      "step": 357
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.308687615526803e-05,
+      "loss": 16.918,
+      "step": 358
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.317929759704251e-05,
+      "loss": 18.7251,
+      "step": 359
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.3271719038817007e-05,
+      "loss": 17.1379,
+      "step": 360
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.33641404805915e-05,
+      "loss": 16.8974,
+      "step": 361
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.345656192236599e-05,
+      "loss": 17.1739,
+      "step": 362
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.354898336414049e-05,
+      "loss": 19.5517,
+      "step": 363
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.364140480591497e-05,
+      "loss": 16.1618,
+      "step": 364
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.3733826247689466e-05,
+      "loss": 17.573,
+      "step": 365
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.3826247689463955e-05,
+      "loss": 17.257,
+      "step": 366
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.391866913123845e-05,
+      "loss": 17.9075,
+      "step": 367
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.401109057301294e-05,
+      "loss": 17.8199,
+      "step": 368
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.410351201478743e-05,
+      "loss": 17.627,
+      "step": 369
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.4195933456561925e-05,
+      "loss": 17.6088,
+      "step": 370
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.4288354898336414e-05,
+      "loss": 15.348,
+      "step": 371
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.438077634011091e-05,
+      "loss": 17.1887,
+      "step": 372
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.44731977818854e-05,
+      "loss": 18.0897,
+      "step": 373
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.456561922365989e-05,
+      "loss": 18.9322,
+      "step": 374
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.4658040665434384e-05,
+      "loss": 15.5636,
+      "step": 375
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.4750462107208874e-05,
+      "loss": 16.827,
+      "step": 376
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.484288354898337e-05,
+      "loss": 19.5084,
+      "step": 377
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.493530499075786e-05,
+      "loss": 16.8976,
+      "step": 378
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.502772643253235e-05,
+      "loss": 17.121,
+      "step": 379
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.5120147874306844e-05,
+      "loss": 19.2101,
+      "step": 380
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.521256931608133e-05,
+      "loss": 17.0868,
+      "step": 381
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.530499075785583e-05,
+      "loss": 17.7931,
+      "step": 382
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.539741219963031e-05,
+      "loss": 19.6144,
+      "step": 383
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.548983364140481e-05,
+      "loss": 19.1626,
+      "step": 384
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.55822550831793e-05,
+      "loss": 16.8115,
+      "step": 385
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.567467652495379e-05,
+      "loss": 17.9764,
+      "step": 386
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.576709796672828e-05,
+      "loss": 17.97,
+      "step": 387
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.585951940850277e-05,
+      "loss": 17.7438,
+      "step": 388
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.5951940850277266e-05,
+      "loss": 16.4815,
+      "step": 389
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.6044362292051755e-05,
+      "loss": 15.1583,
+      "step": 390
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.613678373382625e-05,
+      "loss": 16.8657,
+      "step": 391
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.622920517560074e-05,
+      "loss": 18.8117,
+      "step": 392
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.632162661737523e-05,
+      "loss": 18.893,
+      "step": 393
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.6414048059149726e-05,
+      "loss": 16.3189,
+      "step": 394
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.6506469500924215e-05,
+      "loss": 17.2085,
+      "step": 395
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.659889094269871e-05,
+      "loss": 17.4981,
+      "step": 396
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.66913123844732e-05,
+      "loss": 18.0744,
+      "step": 397
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.678373382624769e-05,
+      "loss": 20.1998,
+      "step": 398
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.6876155268022185e-05,
+      "loss": 19.2819,
+      "step": 399
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.6968576709796674e-05,
+      "loss": 15.5054,
+      "step": 400
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.706099815157117e-05,
+      "loss": 20.9854,
+      "step": 401
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.715341959334566e-05,
+      "loss": 17.6914,
+      "step": 402
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.724584103512015e-05,
+      "loss": 17.8825,
+      "step": 403
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.7338262476894644e-05,
+      "loss": 22.2405,
+      "step": 404
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.743068391866913e-05,
+      "loss": 19.9553,
+      "step": 405
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.752310536044362e-05,
+      "loss": 21.4086,
+      "step": 406
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.761552680221812e-05,
+      "loss": 17.1271,
+      "step": 407
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.770794824399261e-05,
+      "loss": 17.4132,
+      "step": 408
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.78003696857671e-05,
+      "loss": 19.4398,
+      "step": 409
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.789279112754159e-05,
+      "loss": 17.7903,
+      "step": 410
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.798521256931608e-05,
+      "loss": 18.1556,
+      "step": 411
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.807763401109057e-05,
+      "loss": 19.5074,
+      "step": 412
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.817005545286507e-05,
+      "loss": 16.8014,
+      "step": 413
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.826247689463956e-05,
+      "loss": 17.0244,
+      "step": 414
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.835489833641405e-05,
+      "loss": 16.3282,
+      "step": 415
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.844731977818854e-05,
+      "loss": 16.5951,
+      "step": 416
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.853974121996303e-05,
+      "loss": 18.1894,
+      "step": 417
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.8632162661737526e-05,
+      "loss": 18.9452,
+      "step": 418
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.8724584103512015e-05,
+      "loss": 19.9864,
+      "step": 419
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.8817005545286504e-05,
+      "loss": 17.0479,
+      "step": 420
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.8909426987061e-05,
+      "loss": 17.8076,
+      "step": 421
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.900184842883549e-05,
+      "loss": 19.9984,
+      "step": 422
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.9094269870609985e-05,
+      "loss": 18.6048,
+      "step": 423
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.9186691312384474e-05,
+      "loss": 18.9452,
+      "step": 424
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.9279112754158964e-05,
+      "loss": 17.4527,
+      "step": 425
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.937153419593346e-05,
+      "loss": 17.7976,
+      "step": 426
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.946395563770795e-05,
+      "loss": 17.0254,
+      "step": 427
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.9556377079482445e-05,
+      "loss": 16.8357,
+      "step": 428
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.9648798521256934e-05,
+      "loss": 18.5158,
+      "step": 429
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.974121996303142e-05,
+      "loss": 18.4552,
+      "step": 430
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.983364140480592e-05,
+      "loss": 17.9303,
+      "step": 431
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.992606284658041e-05,
+      "loss": 17.4442,
+      "step": 432
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.0018484288354904e-05,
+      "loss": 16.8613,
+      "step": 433
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.011090573012939e-05,
+      "loss": 18.2314,
+      "step": 434
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.020332717190388e-05,
+      "loss": 16.8425,
+      "step": 435
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.029574861367837e-05,
+      "loss": 18.3102,
+      "step": 436
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.038817005545287e-05,
+      "loss": 15.9627,
+      "step": 437
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.048059149722736e-05,
+      "loss": 18.1573,
+      "step": 438
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.0573012939001845e-05,
+      "loss": 15.9093,
+      "step": 439
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.066543438077634e-05,
+      "loss": 17.0638,
+      "step": 440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.075785582255083e-05,
+      "loss": 17.9335,
+      "step": 441
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.0850277264325326e-05,
+      "loss": 20.0075,
+      "step": 442
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.094269870609982e-05,
+      "loss": 17.9466,
+      "step": 443
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.1035120147874305e-05,
+      "loss": 17.2174,
+      "step": 444
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.11275415896488e-05,
+      "loss": 16.0911,
+      "step": 445
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.121996303142329e-05,
+      "loss": 16.1743,
+      "step": 446
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.1312384473197786e-05,
+      "loss": 19.5294,
+      "step": 447
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.1404805914972275e-05,
+      "loss": 20.9506,
+      "step": 448
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.1497227356746764e-05,
+      "loss": 16.718,
+      "step": 449
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.158964879852126e-05,
+      "loss": 17.5032,
+      "step": 450
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.168207024029575e-05,
+      "loss": 16.5786,
+      "step": 451
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.1774491682070245e-05,
+      "loss": 18.4169,
+      "step": 452
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.1866913123844734e-05,
+      "loss": 16.0072,
+      "step": 453
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.195933456561922e-05,
+      "loss": 18.6858,
+      "step": 454
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.205175600739372e-05,
+      "loss": 17.7907,
+      "step": 455
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.214417744916821e-05,
+      "loss": 17.5202,
+      "step": 456
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.2236598890942704e-05,
+      "loss": 17.0853,
+      "step": 457
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.232902033271719e-05,
+      "loss": 16.8813,
+      "step": 458
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.242144177449168e-05,
+      "loss": 18.7169,
+      "step": 459
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.251386321626618e-05,
+      "loss": 18.2961,
+      "step": 460
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.260628465804067e-05,
+      "loss": 17.253,
+      "step": 461
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.2698706099815164e-05,
+      "loss": 14.2926,
+      "step": 462
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.2791127541589646e-05,
+      "loss": 16.958,
+      "step": 463
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.288354898336414e-05,
+      "loss": 19.2609,
+      "step": 464
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.297597042513863e-05,
+      "loss": 18.5109,
+      "step": 465
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.306839186691313e-05,
+      "loss": 17.1921,
+      "step": 466
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.316081330868762e-05,
+      "loss": 18.2002,
+      "step": 467
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.3253234750462105e-05,
+      "loss": 16.6834,
+      "step": 468
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.33456561922366e-05,
+      "loss": 16.7871,
+      "step": 469
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.343807763401109e-05,
+      "loss": 18.2129,
+      "step": 470
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.3530499075785586e-05,
+      "loss": 15.9457,
+      "step": 471
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.3622920517560075e-05,
+      "loss": 17.1247,
+      "step": 472
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.3715341959334564e-05,
+      "loss": 17.6579,
+      "step": 473
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.380776340110906e-05,
+      "loss": 20.4203,
+      "step": 474
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.390018484288355e-05,
+      "loss": 18.4018,
+      "step": 475
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.3992606284658045e-05,
+      "loss": 18.201,
+      "step": 476
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.4085027726432535e-05,
+      "loss": 17.2207,
+      "step": 477
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.4177449168207024e-05,
+      "loss": 17.9318,
+      "step": 478
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.426987060998152e-05,
+      "loss": 17.1236,
+      "step": 479
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.436229205175601e-05,
+      "loss": 16.8856,
+      "step": 480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.4454713493530505e-05,
+      "loss": 14.6961,
+      "step": 481
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.4547134935304994e-05,
+      "loss": 17.1831,
+      "step": 482
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.463955637707948e-05,
+      "loss": 19.9528,
+      "step": 483
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.473197781885398e-05,
+      "loss": 15.6614,
+      "step": 484
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.482439926062847e-05,
+      "loss": 16.3011,
+      "step": 485
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.4916820702402964e-05,
+      "loss": 20.3028,
+      "step": 486
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.5009242144177446e-05,
+      "loss": 17.5163,
+      "step": 487
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.510166358595194e-05,
+      "loss": 19.2545,
+      "step": 488
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.519408502772644e-05,
+      "loss": 17.4618,
+      "step": 489
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.528650646950093e-05,
+      "loss": 16.9832,
+      "step": 490
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.537892791127542e-05,
+      "loss": 18.062,
+      "step": 491
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.5471349353049906e-05,
+      "loss": 17.0362,
+      "step": 492
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.55637707948244e-05,
+      "loss": 19.253,
+      "step": 493
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.565619223659889e-05,
+      "loss": 16.9551,
+      "step": 494
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.5748613678373387e-05,
+      "loss": 16.6881,
+      "step": 495
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.5841035120147876e-05,
+      "loss": 18.9192,
+      "step": 496
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.5933456561922365e-05,
+      "loss": 17.3158,
+      "step": 497
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.602587800369686e-05,
+      "loss": 18.4779,
+      "step": 498
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.611829944547135e-05,
+      "loss": 17.9357,
+      "step": 499
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.6210720887245846e-05,
+      "loss": 18.0473,
+      "step": 500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.6303142329020335e-05,
+      "loss": 20.0578,
+      "step": 501
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.6395563770794824e-05,
+      "loss": 18.6128,
+      "step": 502
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.648798521256932e-05,
+      "loss": 20.8736,
+      "step": 503
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.658040665434381e-05,
+      "loss": 16.7578,
+      "step": 504
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.6672828096118305e-05,
+      "loss": 17.2871,
+      "step": 505
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.6765249537892794e-05,
+      "loss": 18.9869,
+      "step": 506
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.6857670979667283e-05,
+      "loss": 17.5873,
+      "step": 507
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.695009242144178e-05,
+      "loss": 18.2389,
+      "step": 508
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.704251386321627e-05,
+      "loss": 17.3065,
+      "step": 509
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.7134935304990764e-05,
+      "loss": 17.1184,
+      "step": 510
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.722735674676525e-05,
+      "loss": 18.9651,
+      "step": 511
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.731977818853974e-05,
+      "loss": 17.2011,
+      "step": 512
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.741219963031424e-05,
+      "loss": 19.555,
+      "step": 513
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.750462107208873e-05,
+      "loss": 17.7176,
+      "step": 514
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.759704251386322e-05,
+      "loss": 18.6924,
+      "step": 515
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.7689463955637706e-05,
+      "loss": 16.2597,
+      "step": 516
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.77818853974122e-05,
+      "loss": 19.1326,
+      "step": 517
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.787430683918669e-05,
+      "loss": 16.9432,
+      "step": 518
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.796672828096119e-05,
+      "loss": 17.3703,
+      "step": 519
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8059149722735676e-05,
+      "loss": 18.3888,
+      "step": 520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8151571164510165e-05,
+      "loss": 17.5392,
+      "step": 521
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.824399260628466e-05,
+      "loss": 17.9469,
+      "step": 522
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.833641404805915e-05,
+      "loss": 17.2329,
+      "step": 523
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8428835489833646e-05,
+      "loss": 17.045,
+      "step": 524
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8521256931608135e-05,
+      "loss": 18.3625,
+      "step": 525
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8613678373382625e-05,
+      "loss": 17.4384,
+      "step": 526
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.870609981515712e-05,
+      "loss": 16.743,
+      "step": 527
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.879852125693161e-05,
+      "loss": 17.77,
+      "step": 528
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8890942698706106e-05,
+      "loss": 19.5948,
+      "step": 529
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8983364140480595e-05,
+      "loss": 20.682,
+      "step": 530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9075785582255084e-05,
+      "loss": 20.5962,
+      "step": 531
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.916820702402958e-05,
+      "loss": 16.2137,
+      "step": 532
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.926062846580407e-05,
+      "loss": 15.2998,
+      "step": 533
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.935304990757856e-05,
+      "loss": 19.0108,
+      "step": 534
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9445471349353054e-05,
+      "loss": 16.6598,
+      "step": 535
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.953789279112754e-05,
+      "loss": 19.4124,
+      "step": 536
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.963031423290204e-05,
+      "loss": 16.7912,
+      "step": 537
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.972273567467653e-05,
+      "loss": 17.4258,
+      "step": 538
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.981515711645102e-05,
+      "loss": 17.9191,
+      "step": 539
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9907578558225506e-05,
+      "loss": 19.4315,
+      "step": 540
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5e-05,
+      "loss": 19.1898,
+      "step": 541
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.00924214417745e-05,
+      "loss": 18.7414,
+      "step": 542
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.018484288354899e-05,
+      "loss": 15.5225,
+      "step": 543
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.027726432532348e-05,
+      "loss": 16.8118,
+      "step": 544
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.036968576709797e-05,
+      "loss": 16.9417,
+      "step": 545
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.0462107208872455e-05,
+      "loss": 17.4174,
+      "step": 546
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.055452865064695e-05,
+      "loss": 17.4082,
+      "step": 547
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.064695009242144e-05,
+      "loss": 17.3384,
+      "step": 548
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.0739371534195936e-05,
+      "loss": 17.8814,
+      "step": 549
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.0831792975970425e-05,
+      "loss": 18.4522,
+      "step": 550
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.092421441774492e-05,
+      "loss": 17.8547,
+      "step": 551
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.101663585951941e-05,
+      "loss": 18.5363,
+      "step": 552
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.1109057301293906e-05,
+      "loss": 16.3083,
+      "step": 553
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.1201478743068395e-05,
+      "loss": 18.0965,
+      "step": 554
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.129390018484289e-05,
+      "loss": 16.72,
+      "step": 555
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.1386321626617373e-05,
+      "loss": 16.8634,
+      "step": 556
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.147874306839186e-05,
+      "loss": 17.8812,
+      "step": 557
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.157116451016636e-05,
+      "loss": 20.4323,
+      "step": 558
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.1663585951940854e-05,
+      "loss": 18.4533,
+      "step": 559
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.1756007393715344e-05,
+      "loss": 18.0217,
+      "step": 560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.184842883548984e-05,
+      "loss": 17.7003,
+      "step": 561
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.194085027726433e-05,
+      "loss": 15.9481,
+      "step": 562
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.2033271719038825e-05,
+      "loss": 17.9688,
+      "step": 563
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.2125693160813314e-05,
+      "loss": 18.95,
+      "step": 564
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.2218114602587796e-05,
+      "loss": 16.51,
+      "step": 565
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.231053604436229e-05,
+      "loss": 18.3397,
+      "step": 566
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.240295748613678e-05,
+      "loss": 16.6805,
+      "step": 567
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.249537892791128e-05,
+      "loss": 18.4973,
+      "step": 568
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.2587800369685766e-05,
+      "loss": 17.4778,
+      "step": 569
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.268022181146026e-05,
+      "loss": 16.7314,
+      "step": 570
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.277264325323476e-05,
+      "loss": 17.9001,
+      "step": 571
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.286506469500925e-05,
+      "loss": 17.8572,
+      "step": 572
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.295748613678374e-05,
+      "loss": 19.7544,
+      "step": 573
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.304990757855823e-05,
+      "loss": 17.4575,
+      "step": 574
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.3142329020332715e-05,
+      "loss": 20.7827,
+      "step": 575
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.323475046210721e-05,
+      "loss": 19.2229,
+      "step": 576
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.33271719038817e-05,
+      "loss": 17.2744,
+      "step": 577
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.3419593345656196e-05,
+      "loss": 17.8176,
+      "step": 578
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.3512014787430685e-05,
+      "loss": 17.8874,
+      "step": 579
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.360443622920518e-05,
+      "loss": 20.1579,
+      "step": 580
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.369685767097967e-05,
+      "loss": 18.7735,
+      "step": 581
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.3789279112754166e-05,
+      "loss": 18.1399,
+      "step": 582
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.3881700554528655e-05,
+      "loss": 18.0929,
+      "step": 583
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.397412199630314e-05,
+      "loss": 17.4217,
+      "step": 584
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.406654343807763e-05,
+      "loss": 17.4558,
+      "step": 585
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.415896487985212e-05,
+      "loss": 14.9792,
+      "step": 586
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.425138632162662e-05,
+      "loss": 15.621,
+      "step": 587
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.4343807763401114e-05,
+      "loss": 17.9243,
+      "step": 588
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.44362292051756e-05,
+      "loss": 17.8779,
+      "step": 589
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.45286506469501e-05,
+      "loss": 18.1541,
+      "step": 590
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.462107208872459e-05,
+      "loss": 17.5245,
+      "step": 591
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.4713493530499084e-05,
+      "loss": 16.8786,
+      "step": 592
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.4805914972273573e-05,
+      "loss": 16.1402,
+      "step": 593
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.4898336414048056e-05,
+      "loss": 18.853,
+      "step": 594
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.499075785582255e-05,
+      "loss": 20.1003,
+      "step": 595
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.508317929759704e-05,
+      "loss": 17.8899,
+      "step": 596
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.517560073937154e-05,
+      "loss": 18.6151,
+      "step": 597
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.5268022181146026e-05,
+      "loss": 16.8337,
+      "step": 598
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.536044362292052e-05,
+      "loss": 17.3565,
+      "step": 599
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.545286506469501e-05,
+      "loss": 18.1815,
+      "step": 600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.554528650646951e-05,
+      "loss": 17.1634,
+      "step": 601
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.5637707948244e-05,
+      "loss": 17.669,
+      "step": 602
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.573012939001848e-05,
+      "loss": 18.5952,
+      "step": 603
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.5822550831792974e-05,
+      "loss": 16.1357,
+      "step": 604
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.591497227356747e-05,
+      "loss": 18.5698,
+      "step": 605
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.600739371534196e-05,
+      "loss": 16.5488,
+      "step": 606
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.6099815157116455e-05,
+      "loss": 18.4043,
+      "step": 607
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.6192236598890944e-05,
+      "loss": 20.5244,
+      "step": 608
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.628465804066544e-05,
+      "loss": 16.3769,
+      "step": 609
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.637707948243993e-05,
+      "loss": 17.0118,
+      "step": 610
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.6469500924214425e-05,
+      "loss": 18.413,
+      "step": 611
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.6561922365988915e-05,
+      "loss": 19.3694,
+      "step": 612
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.66543438077634e-05,
+      "loss": 17.0299,
+      "step": 613
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.674676524953789e-05,
+      "loss": 18.6144,
+      "step": 614
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.683918669131238e-05,
+      "loss": 19.1739,
+      "step": 615
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.693160813308688e-05,
+      "loss": 18.478,
+      "step": 616
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.7024029574861374e-05,
+      "loss": 16.892,
+      "step": 617
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.711645101663586e-05,
+      "loss": 18.4209,
+      "step": 618
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.720887245841036e-05,
+      "loss": 17.5796,
+      "step": 619
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.730129390018485e-05,
+      "loss": 19.0904,
+      "step": 620
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.7393715341959344e-05,
+      "loss": 16.9095,
+      "step": 621
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.7486136783733826e-05,
+      "loss": 15.9882,
+      "step": 622
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.7578558225508315e-05,
+      "loss": 18.0468,
+      "step": 623
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.767097966728281e-05,
+      "loss": 19.2096,
+      "step": 624
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.77634011090573e-05,
+      "loss": 17.4982,
+      "step": 625
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.7855822550831796e-05,
+      "loss": 16.986,
+      "step": 626
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.7948243992606286e-05,
+      "loss": 17.7198,
+      "step": 627
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.804066543438078e-05,
+      "loss": 19.6353,
+      "step": 628
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.813308687615527e-05,
+      "loss": 16.6337,
+      "step": 629
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.8225508317929767e-05,
+      "loss": 20.2217,
+      "step": 630
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.831792975970426e-05,
+      "loss": 18.2715,
+      "step": 631
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.841035120147874e-05,
+      "loss": 20.1764,
+      "step": 632
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.8502772643253234e-05,
+      "loss": 19.0841,
+      "step": 633
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.859519408502773e-05,
+      "loss": 17.5812,
+      "step": 634
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.868761552680222e-05,
+      "loss": 17.0779,
+      "step": 635
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.8780036968576715e-05,
+      "loss": 16.9006,
+      "step": 636
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.8872458410351204e-05,
+      "loss": 19.7057,
+      "step": 637
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.89648798521257e-05,
+      "loss": 15.5787,
+      "step": 638
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.905730129390019e-05,
+      "loss": 16.6821,
+      "step": 639
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.9149722735674685e-05,
+      "loss": 18.4224,
+      "step": 640
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.924214417744917e-05,
+      "loss": 17.8699,
+      "step": 641
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.933456561922366e-05,
+      "loss": 17.8386,
+      "step": 642
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.942698706099815e-05,
+      "loss": 17.1367,
+      "step": 643
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.951940850277264e-05,
+      "loss": 18.7376,
+      "step": 644
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.961182994454714e-05,
+      "loss": 18.559,
+      "step": 645
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.9704251386321634e-05,
+      "loss": 16.1036,
+      "step": 646
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.979667282809612e-05,
+      "loss": 18.8402,
+      "step": 647
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.988909426987062e-05,
+      "loss": 17.7071,
+      "step": 648
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.998151571164511e-05,
+      "loss": 17.2706,
+      "step": 649
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.0073937153419604e-05,
+      "loss": 18.4882,
+      "step": 650
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.0166358595194086e-05,
+      "loss": 18.0344,
+      "step": 651
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.0258780036968575e-05,
+      "loss": 18.2904,
+      "step": 652
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.035120147874307e-05,
+      "loss": 16.6123,
+      "step": 653
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.044362292051756e-05,
+      "loss": 16.6194,
+      "step": 654
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.0536044362292056e-05,
+      "loss": 17.903,
+      "step": 655
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.0628465804066545e-05,
+      "loss": 19.1856,
+      "step": 656
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.072088724584104e-05,
+      "loss": 18.0237,
+      "step": 657
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.081330868761553e-05,
+      "loss": 16.3277,
+      "step": 658
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.0905730129390026e-05,
+      "loss": 16.9367,
+      "step": 659
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.099815157116451e-05,
+      "loss": 17.8714,
+      "step": 660
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.1090573012939e-05,
+      "loss": 16.6708,
+      "step": 661
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.11829944547135e-05,
+      "loss": 17.7274,
+      "step": 662
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.127541589648798e-05,
+      "loss": 17.1512,
+      "step": 663
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.136783733826249e-05,
+      "loss": 17.5843,
+      "step": 664
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.146025878003697e-05,
+      "loss": 15.9626,
+      "step": 665
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.155268022181146e-05,
+      "loss": 16.7436,
+      "step": 666
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.164510166358595e-05,
+      "loss": 16.8996,
+      "step": 667
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.173752310536046e-05,
+      "loss": 16.0679,
+      "step": 668
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.182994454713494e-05,
+      "loss": 18.8146,
+      "step": 669
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.192236598890942e-05,
+      "loss": 18.3977,
+      "step": 670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.201478743068392e-05,
+      "loss": 18.9217,
+      "step": 671
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.210720887245841e-05,
+      "loss": 18.9486,
+      "step": 672
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.21996303142329e-05,
+      "loss": 17.9355,
+      "step": 673
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.229205175600739e-05,
+      "loss": 17.1283,
+      "step": 674
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.23844731977819e-05,
+      "loss": 18.8216,
+      "step": 675
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.247689463955638e-05,
+      "loss": 16.981,
+      "step": 676
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.256931608133087e-05,
+      "loss": 17.0277,
+      "step": 677
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.266173752310537e-05,
+      "loss": 18.6495,
+      "step": 678
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.275415896487985e-05,
+      "loss": 18.3273,
+      "step": 679
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.284658040665434e-05,
+      "loss": 18.3467,
+      "step": 680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.293900184842884e-05,
+      "loss": 16.5232,
+      "step": 681
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.303142329020333e-05,
+      "loss": 16.8944,
+      "step": 682
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.312384473197782e-05,
+      "loss": 18.3669,
+      "step": 683
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.321626617375231e-05,
+      "loss": 17.4527,
+      "step": 684
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.330868761552681e-05,
+      "loss": 18.675,
+      "step": 685
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.34011090573013e-05,
+      "loss": 18.5863,
+      "step": 686
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.349353049907579e-05,
+      "loss": 18.2625,
+      "step": 687
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.358595194085028e-05,
+      "loss": 19.3844,
+      "step": 688
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.367837338262477e-05,
+      "loss": 17.9848,
+      "step": 689
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.377079482439926e-05,
+      "loss": 18.4912,
+      "step": 690
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.386321626617375e-05,
+      "loss": 18.5247,
+      "step": 691
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.395563770794825e-05,
+      "loss": 17.2853,
+      "step": 692
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.404805914972274e-05,
+      "loss": 17.8395,
+      "step": 693
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.414048059149723e-05,
+      "loss": 17.6271,
+      "step": 694
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.423290203327173e-05,
+      "loss": 16.8879,
+      "step": 695
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.432532347504622e-05,
+      "loss": 15.2856,
+      "step": 696
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.441774491682071e-05,
+      "loss": 16.2758,
+      "step": 697
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.45101663585952e-05,
+      "loss": 18.8332,
+      "step": 698
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.460258780036969e-05,
+      "loss": 16.5944,
+      "step": 699
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.469500924214418e-05,
+      "loss": 18.7502,
+      "step": 700
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.478743068391867e-05,
+      "loss": 19.4832,
+      "step": 701
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.487985212569317e-05,
+      "loss": 18.3129,
+      "step": 702
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.497227356746766e-05,
+      "loss": 19.3043,
+      "step": 703
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.506469500924215e-05,
+      "loss": 17.1846,
+      "step": 704
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.515711645101664e-05,
+      "loss": 16.8566,
+      "step": 705
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.524953789279114e-05,
+      "loss": 17.5874,
+      "step": 706
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.534195933456563e-05,
+      "loss": 18.2839,
+      "step": 707
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.543438077634012e-05,
+      "loss": 16.5492,
+      "step": 708
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.55268022181146e-05,
+      "loss": 16.763,
+      "step": 709
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.56192236598891e-05,
+      "loss": 19.1905,
+      "step": 710
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.571164510166358e-05,
+      "loss": 19.3873,
+      "step": 711
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.580406654343809e-05,
+      "loss": 19.0397,
+      "step": 712
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.589648798521258e-05,
+      "loss": 18.2267,
+      "step": 713
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.598890942698706e-05,
+      "loss": 16.7844,
+      "step": 714
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.608133086876155e-05,
+      "loss": 19.3658,
+      "step": 715
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.617375231053606e-05,
+      "loss": 17.8323,
+      "step": 716
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.626617375231053e-05,
+      "loss": 15.7578,
+      "step": 717
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.635859519408502e-05,
+      "loss": 17.6948,
+      "step": 718
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.645101663585952e-05,
+      "loss": 20.014,
+      "step": 719
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.654343807763401e-05,
+      "loss": 17.7379,
+      "step": 720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.66358595194085e-05,
+      "loss": 19.524,
+      "step": 721
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.6728280961183e-05,
+      "loss": 19.0029,
+      "step": 722
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.68207024029575e-05,
+      "loss": 19.45,
+      "step": 723
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.691312384473198e-05,
+      "loss": 19.3666,
+      "step": 724
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.700554528650647e-05,
+      "loss": 18.831,
+      "step": 725
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.709796672828098e-05,
+      "loss": 16.4688,
+      "step": 726
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.719038817005545e-05,
+      "loss": 19.2285,
+      "step": 727
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.728280961182994e-05,
+      "loss": 18.0891,
+      "step": 728
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.737523105360444e-05,
+      "loss": 18.3484,
+      "step": 729
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.746765249537893e-05,
+      "loss": 17.5895,
+      "step": 730
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.756007393715342e-05,
+      "loss": 16.9859,
+      "step": 731
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.765249537892791e-05,
+      "loss": 18.3135,
+      "step": 732
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.774491682070241e-05,
+      "loss": 18.6734,
+      "step": 733
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.78373382624769e-05,
+      "loss": 16.6561,
+      "step": 734
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.792975970425139e-05,
+      "loss": 17.2632,
+      "step": 735
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.802218114602588e-05,
+      "loss": 15.7814,
+      "step": 736
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.811460258780037e-05,
+      "loss": 18.7512,
+      "step": 737
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.820702402957486e-05,
+      "loss": 19.4107,
+      "step": 738
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.829944547134936e-05,
+      "loss": 18.3001,
+      "step": 739
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.839186691312385e-05,
+      "loss": 18.7987,
+      "step": 740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.848428835489834e-05,
+      "loss": 15.6542,
+      "step": 741
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.857670979667283e-05,
+      "loss": 19.0048,
+      "step": 742
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.866913123844733e-05,
+      "loss": 17.3055,
+      "step": 743
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.876155268022182e-05,
+      "loss": 17.6123,
+      "step": 744
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.885397412199631e-05,
+      "loss": 16.6975,
+      "step": 745
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.89463955637708e-05,
+      "loss": 18.3547,
+      "step": 746
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.903881700554529e-05,
+      "loss": 18.7491,
+      "step": 747
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.913123844731978e-05,
+      "loss": 18.514,
+      "step": 748
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.922365988909427e-05,
+      "loss": 17.8367,
+      "step": 749
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.931608133086877e-05,
+      "loss": 17.6921,
+      "step": 750
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.940850277264326e-05,
+      "loss": 17.653,
+      "step": 751
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.950092421441775e-05,
+      "loss": 18.6252,
+      "step": 752
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.959334565619225e-05,
+      "loss": 18.0083,
+      "step": 753
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.968576709796674e-05,
+      "loss": 17.6265,
+      "step": 754
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.977818853974121e-05,
+      "loss": 17.0541,
+      "step": 755
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.987060998151572e-05,
+      "loss": 17.1256,
+      "step": 756
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 6.99630314232902e-05,
+      "loss": 17.0798,
+      "step": 757
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.00554528650647e-05,
+      "loss": 18.3368,
+      "step": 758
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.014787430683918e-05,
+      "loss": 18.1764,
+      "step": 759
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.024029574861369e-05,
+      "loss": 18.4486,
+      "step": 760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.033271719038818e-05,
+      "loss": 16.7201,
+      "step": 761
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.042513863216267e-05,
+      "loss": 18.7751,
+      "step": 762
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.051756007393715e-05,
+      "loss": 17.45,
+      "step": 763
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.060998151571166e-05,
+      "loss": 18.2407,
+      "step": 764
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.070240295748613e-05,
+      "loss": 18.6891,
+      "step": 765
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.079482439926062e-05,
+      "loss": 19.7776,
+      "step": 766
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.088724584103512e-05,
+      "loss": 17.1263,
+      "step": 767
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.097966728280961e-05,
+      "loss": 19.4172,
+      "step": 768
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.10720887245841e-05,
+      "loss": 18.5818,
+      "step": 769
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.11645101663586e-05,
+      "loss": 17.6584,
+      "step": 770
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.12569316081331e-05,
+      "loss": 17.7316,
+      "step": 771
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.134935304990758e-05,
+      "loss": 16.7923,
+      "step": 772
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.144177449168207e-05,
+      "loss": 18.2307,
+      "step": 773
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.153419593345656e-05,
+      "loss": 18.6822,
+      "step": 774
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.162661737523105e-05,
+      "loss": 17.8764,
+      "step": 775
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.171903881700554e-05,
+      "loss": 18.5043,
+      "step": 776
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.181146025878004e-05,
+      "loss": 19.6225,
+      "step": 777
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.190388170055453e-05,
+      "loss": 17.9581,
+      "step": 778
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.199630314232902e-05,
+      "loss": 17.7672,
+      "step": 779
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.208872458410351e-05,
+      "loss": 19.0525,
+      "step": 780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.218114602587801e-05,
+      "loss": 16.0974,
+      "step": 781
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.22735674676525e-05,
+      "loss": 20.3029,
+      "step": 782
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.236598890942699e-05,
+      "loss": 17.595,
+      "step": 783
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.245841035120148e-05,
+      "loss": 17.7613,
+      "step": 784
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.255083179297597e-05,
+      "loss": 16.5795,
+      "step": 785
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.264325323475046e-05,
+      "loss": 14.7767,
+      "step": 786
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.273567467652496e-05,
+      "loss": 17.6748,
+      "step": 787
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.282809611829945e-05,
+      "loss": 17.1199,
+      "step": 788
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.292051756007394e-05,
+      "loss": 18.8401,
+      "step": 789
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.301293900184843e-05,
+      "loss": 17.4914,
+      "step": 790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.310536044362293e-05,
+      "loss": 17.5229,
+      "step": 791
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.319778188539742e-05,
+      "loss": 20.5858,
+      "step": 792
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.32902033271719e-05,
+      "loss": 17.9247,
+      "step": 793
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.33826247689464e-05,
+      "loss": 16.4542,
+      "step": 794
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.347504621072089e-05,
+      "loss": 18.6346,
+      "step": 795
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.356746765249538e-05,
+      "loss": 17.379,
+      "step": 796
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.365988909426988e-05,
+      "loss": 16.556,
+      "step": 797
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.375231053604437e-05,
+      "loss": 16.4406,
+      "step": 798
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.384473197781886e-05,
+      "loss": 17.6065,
+      "step": 799
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.393715341959335e-05,
+      "loss": 17.9225,
+      "step": 800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.402957486136785e-05,
+      "loss": 17.2447,
+      "step": 801
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.412199630314234e-05,
+      "loss": 15.3461,
+      "step": 802
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.421441774491682e-05,
+      "loss": 18.2239,
+      "step": 803
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.430683918669132e-05,
+      "loss": 20.5382,
+      "step": 804
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.439926062846581e-05,
+      "loss": 17.8141,
+      "step": 805
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.44916820702403e-05,
+      "loss": 19.4097,
+      "step": 806
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.458410351201479e-05,
+      "loss": 20.756,
+      "step": 807
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.467652495378929e-05,
+      "loss": 18.4039,
+      "step": 808
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.476894639556378e-05,
+      "loss": 18.1873,
+      "step": 809
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.486136783733827e-05,
+      "loss": 16.8122,
+      "step": 810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.495378927911277e-05,
+      "loss": 17.6913,
+      "step": 811
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.504621072088724e-05,
+      "loss": 15.4485,
+      "step": 812
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.513863216266173e-05,
+      "loss": 16.3507,
+      "step": 813
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.523105360443624e-05,
+      "loss": 18.2226,
+      "step": 814
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.532347504621073e-05,
+      "loss": 18.8782,
+      "step": 815
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.541589648798521e-05,
+      "loss": 18.9261,
+      "step": 816
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.55083179297597e-05,
+      "loss": 17.0652,
+      "step": 817
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.56007393715342e-05,
+      "loss": 19.8919,
+      "step": 818
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.56931608133087e-05,
+      "loss": 18.0075,
+      "step": 819
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.578558225508319e-05,
+      "loss": 17.5506,
+      "step": 820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.587800369685767e-05,
+      "loss": 17.5679,
+      "step": 821
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.597042513863216e-05,
+      "loss": 17.1589,
+      "step": 822
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.606284658040665e-05,
+      "loss": 19.0194,
+      "step": 823
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.615526802218114e-05,
+      "loss": 19.385,
+      "step": 824
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.624768946395564e-05,
+      "loss": 18.4109,
+      "step": 825
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.634011090573013e-05,
+      "loss": 19.1603,
+      "step": 826
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.643253234750462e-05,
+      "loss": 16.2443,
+      "step": 827
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.652495378927913e-05,
+      "loss": 17.0007,
+      "step": 828
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.661737523105361e-05,
+      "loss": 17.5691,
+      "step": 829
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.67097966728281e-05,
+      "loss": 19.4221,
+      "step": 830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.680221811460259e-05,
+      "loss": 17.4324,
+      "step": 831
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.689463955637708e-05,
+      "loss": 18.8888,
+      "step": 832
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.698706099815157e-05,
+      "loss": 17.4898,
+      "step": 833
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.707948243992606e-05,
+      "loss": 17.0232,
+      "step": 834
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.717190388170056e-05,
+      "loss": 17.738,
+      "step": 835
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.726432532347505e-05,
+      "loss": 15.9859,
+      "step": 836
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.735674676524954e-05,
+      "loss": 16.7543,
+      "step": 837
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.744916820702403e-05,
+      "loss": 15.576,
+      "step": 838
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.754158964879853e-05,
+      "loss": 17.7202,
+      "step": 839
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.763401109057301e-05,
+      "loss": 20.7313,
+      "step": 840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.77264325323475e-05,
+      "loss": 15.969,
+      "step": 841
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.7818853974122e-05,
+      "loss": 17.0533,
+      "step": 842
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.791127541589649e-05,
+      "loss": 20.7243,
+      "step": 843
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.800369685767098e-05,
+      "loss": 18.7093,
+      "step": 844
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.809611829944548e-05,
+      "loss": 17.3398,
+      "step": 845
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.818853974121997e-05,
+      "loss": 16.4045,
+      "step": 846
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.828096118299446e-05,
+      "loss": 14.9688,
+      "step": 847
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.837338262476895e-05,
+      "loss": 18.6574,
+      "step": 848
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.846580406654345e-05,
+      "loss": 16.9957,
+      "step": 849
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.855822550831793e-05,
+      "loss": 16.9835,
+      "step": 850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.865064695009242e-05,
+      "loss": 17.3747,
+      "step": 851
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.874306839186692e-05,
+      "loss": 18.2215,
+      "step": 852
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.883548983364141e-05,
+      "loss": 18.8867,
+      "step": 853
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.89279112754159e-05,
+      "loss": 18.4429,
+      "step": 854
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.902033271719039e-05,
+      "loss": 19.0353,
+      "step": 855
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.911275415896489e-05,
+      "loss": 17.8078,
+      "step": 856
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.920517560073938e-05,
+      "loss": 17.7048,
+      "step": 857
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.929759704251387e-05,
+      "loss": 16.927,
+      "step": 858
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.939001848428836e-05,
+      "loss": 14.8584,
+      "step": 859
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.948243992606285e-05,
+      "loss": 18.1606,
+      "step": 860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.957486136783733e-05,
+      "loss": 15.9139,
+      "step": 861
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.966728280961184e-05,
+      "loss": 15.4887,
+      "step": 862
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.975970425138633e-05,
+      "loss": 19.8994,
+      "step": 863
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.985212569316082e-05,
+      "loss": 17.4391,
+      "step": 864
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.99445471349353e-05,
+      "loss": 17.0895,
+      "step": 865
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.003696857670981e-05,
+      "loss": 17.8374,
+      "step": 866
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.01293900184843e-05,
+      "loss": 16.9721,
+      "step": 867
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.022181146025879e-05,
+      "loss": 17.4934,
+      "step": 868
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.031423290203328e-05,
+      "loss": 15.5529,
+      "step": 869
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.040665434380776e-05,
+      "loss": 17.8946,
+      "step": 870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.049907578558225e-05,
+      "loss": 17.3465,
+      "step": 871
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.059149722735674e-05,
+      "loss": 17.378,
+      "step": 872
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.068391866913125e-05,
+      "loss": 16.8185,
+      "step": 873
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.077634011090573e-05,
+      "loss": 17.4457,
+      "step": 874
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.086876155268022e-05,
+      "loss": 18.416,
+      "step": 875
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.096118299445473e-05,
+      "loss": 16.8341,
+      "step": 876
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.105360443622922e-05,
+      "loss": 19.585,
+      "step": 877
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.114602587800369e-05,
+      "loss": 17.4264,
+      "step": 878
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.12384473197782e-05,
+      "loss": 17.1488,
+      "step": 879
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.133086876155268e-05,
+      "loss": 16.0877,
+      "step": 880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.142329020332717e-05,
+      "loss": 18.9064,
+      "step": 881
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.151571164510166e-05,
+      "loss": 17.3865,
+      "step": 882
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.160813308687616e-05,
+      "loss": 17.2017,
+      "step": 883
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.170055452865065e-05,
+      "loss": 17.2138,
+      "step": 884
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.179297597042514e-05,
+      "loss": 14.6404,
+      "step": 885
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.188539741219964e-05,
+      "loss": 19.3862,
+      "step": 886
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.197781885397413e-05,
+      "loss": 17.6125,
+      "step": 887
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.207024029574861e-05,
+      "loss": 17.3617,
+      "step": 888
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.216266173752311e-05,
+      "loss": 17.379,
+      "step": 889
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.22550831792976e-05,
+      "loss": 19.1444,
+      "step": 890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.234750462107209e-05,
+      "loss": 17.0935,
+      "step": 891
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.243992606284658e-05,
+      "loss": 18.3134,
+      "step": 892
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.253234750462108e-05,
+      "loss": 18.1902,
+      "step": 893
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.262476894639557e-05,
+      "loss": 19.3797,
+      "step": 894
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.271719038817006e-05,
+      "loss": 17.6493,
+      "step": 895
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.280961182994455e-05,
+      "loss": 17.3755,
+      "step": 896
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.290203327171904e-05,
+      "loss": 17.8788,
+      "step": 897
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.299445471349353e-05,
+      "loss": 17.325,
+      "step": 898
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.308687615526802e-05,
+      "loss": 18.9666,
+      "step": 899
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.317929759704252e-05,
+      "loss": 16.6433,
+      "step": 900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.327171903881701e-05,
+      "loss": 19.3607,
+      "step": 901
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.33641404805915e-05,
+      "loss": 17.4182,
+      "step": 902
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.3456561922366e-05,
+      "loss": 16.7147,
+      "step": 903
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.354898336414049e-05,
+      "loss": 17.4002,
+      "step": 904
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.364140480591498e-05,
+      "loss": 19.3869,
+      "step": 905
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.373382624768947e-05,
+      "loss": 15.7413,
+      "step": 906
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.382624768946396e-05,
+      "loss": 18.1469,
+      "step": 907
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.391866913123845e-05,
+      "loss": 18.9388,
+      "step": 908
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.401109057301294e-05,
+      "loss": 19.8349,
+      "step": 909
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.410351201478744e-05,
+      "loss": 16.1967,
+      "step": 910
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.419593345656193e-05,
+      "loss": 19.3354,
+      "step": 911
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.428835489833642e-05,
+      "loss": 17.817,
+      "step": 912
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.43807763401109e-05,
+      "loss": 20.207,
+      "step": 913
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.447319778188541e-05,
+      "loss": 21.3498,
+      "step": 914
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.45656192236599e-05,
+      "loss": 15.232,
+      "step": 915
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.465804066543437e-05,
+      "loss": 17.7416,
+      "step": 916
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.475046210720888e-05,
+      "loss": 17.4676,
+      "step": 917
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.484288354898337e-05,
+      "loss": 17.5387,
+      "step": 918
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.493530499075785e-05,
+      "loss": 16.4941,
+      "step": 919
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.502772643253236e-05,
+      "loss": 17.2474,
+      "step": 920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.512014787430685e-05,
+      "loss": 17.8668,
+      "step": 921
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.521256931608134e-05,
+      "loss": 17.1858,
+      "step": 922
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.530499075785582e-05,
+      "loss": 17.8292,
+      "step": 923
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.539741219963033e-05,
+      "loss": 17.6384,
+      "step": 924
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.548983364140482e-05,
+      "loss": 16.1412,
+      "step": 925
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.558225508317929e-05,
+      "loss": 18.3211,
+      "step": 926
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.56746765249538e-05,
+      "loss": 18.8804,
+      "step": 927
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.576709796672828e-05,
+      "loss": 17.977,
+      "step": 928
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.585951940850277e-05,
+      "loss": 19.2902,
+      "step": 929
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.595194085027726e-05,
+      "loss": 15.5813,
+      "step": 930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.604436229205176e-05,
+      "loss": 16.2296,
+      "step": 931
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.613678373382625e-05,
+      "loss": 18.4856,
+      "step": 932
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.622920517560074e-05,
+      "loss": 20.8275,
+      "step": 933
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.632162661737525e-05,
+      "loss": 15.42,
+      "step": 934
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.641404805914972e-05,
+      "loss": 17.0737,
+      "step": 935
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.650646950092421e-05,
+      "loss": 17.0592,
+      "step": 936
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.659889094269871e-05,
+      "loss": 17.7171,
+      "step": 937
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.66913123844732e-05,
+      "loss": 18.8775,
+      "step": 938
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.678373382624769e-05,
+      "loss": 18.8069,
+      "step": 939
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.687615526802218e-05,
+      "loss": 19.9411,
+      "step": 940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.696857670979668e-05,
+      "loss": 18.8081,
+      "step": 941
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.706099815157117e-05,
+      "loss": 19.6248,
+      "step": 942
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.715341959334566e-05,
+      "loss": 19.2902,
+      "step": 943
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.724584103512015e-05,
+      "loss": 18.9085,
+      "step": 944
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.733826247689464e-05,
+      "loss": 19.2566,
+      "step": 945
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.743068391866913e-05,
+      "loss": 20.8487,
+      "step": 946
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.752310536044362e-05,
+      "loss": 18.6409,
+      "step": 947
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.761552680221812e-05,
+      "loss": 18.0967,
+      "step": 948
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.770794824399261e-05,
+      "loss": 18.7634,
+      "step": 949
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.78003696857671e-05,
+      "loss": 18.3144,
+      "step": 950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.78927911275416e-05,
+      "loss": 16.776,
+      "step": 951
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.798521256931609e-05,
+      "loss": 19.9267,
+      "step": 952
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.807763401109058e-05,
+      "loss": 16.8589,
+      "step": 953
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.817005545286507e-05,
+      "loss": 17.6802,
+      "step": 954
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.826247689463956e-05,
+      "loss": 17.345,
+      "step": 955
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.835489833641405e-05,
+      "loss": 18.6041,
+      "step": 956
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.844731977818854e-05,
+      "loss": 15.6561,
+      "step": 957
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.853974121996304e-05,
+      "loss": 17.4328,
+      "step": 958
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.863216266173753e-05,
+      "loss": 16.6355,
+      "step": 959
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.872458410351202e-05,
+      "loss": 16.779,
+      "step": 960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.881700554528652e-05,
+      "loss": 18.3982,
+      "step": 961
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.890942698706101e-05,
+      "loss": 16.5074,
+      "step": 962
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.90018484288355e-05,
+      "loss": 16.4963,
+      "step": 963
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.909426987060999e-05,
+      "loss": 16.1621,
+      "step": 964
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.918669131238448e-05,
+      "loss": 18.5709,
+      "step": 965
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.927911275415897e-05,
+      "loss": 20.7659,
+      "step": 966
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.937153419593346e-05,
+      "loss": 20.3466,
+      "step": 967
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.946395563770796e-05,
+      "loss": 17.7923,
+      "step": 968
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.955637707948245e-05,
+      "loss": 18.6931,
+      "step": 969
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.964879852125694e-05,
+      "loss": 17.0742,
+      "step": 970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.974121996303143e-05,
+      "loss": 16.9079,
+      "step": 971
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.983364140480593e-05,
+      "loss": 16.7328,
+      "step": 972
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.99260628465804e-05,
+      "loss": 17.2646,
+      "step": 973
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.001848428835489e-05,
+      "loss": 17.3825,
+      "step": 974
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.01109057301294e-05,
+      "loss": 16.9011,
+      "step": 975
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.020332717190388e-05,
+      "loss": 15.856,
+      "step": 976
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.029574861367837e-05,
+      "loss": 17.6341,
+      "step": 977
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.038817005545288e-05,
+      "loss": 18.903,
+      "step": 978
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.048059149722737e-05,
+      "loss": 16.4374,
+      "step": 979
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.057301293900185e-05,
+      "loss": 18.4854,
+      "step": 980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.066543438077634e-05,
+      "loss": 16.6037,
+      "step": 981
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.075785582255085e-05,
+      "loss": 17.1322,
+      "step": 982
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.085027726432532e-05,
+      "loss": 17.4951,
+      "step": 983
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.094269870609981e-05,
+      "loss": 16.5681,
+      "step": 984
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.103512014787431e-05,
+      "loss": 18.4174,
+      "step": 985
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.11275415896488e-05,
+      "loss": 18.0036,
+      "step": 986
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.121996303142329e-05,
+      "loss": 17.3391,
+      "step": 987
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.131238447319778e-05,
+      "loss": 16.5487,
+      "step": 988
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.140480591497228e-05,
+      "loss": 17.9172,
+      "step": 989
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.149722735674677e-05,
+      "loss": 19.0572,
+      "step": 990
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.158964879852126e-05,
+      "loss": 16.8275,
+      "step": 991
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.168207024029575e-05,
+      "loss": 18.4342,
+      "step": 992
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.177449168207024e-05,
+      "loss": 16.853,
+      "step": 993
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.186691312384473e-05,
+      "loss": 17.2107,
+      "step": 994
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.195933456561923e-05,
+      "loss": 17.0466,
+      "step": 995
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.205175600739372e-05,
+      "loss": 17.2062,
+      "step": 996
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.214417744916821e-05,
+      "loss": 18.1825,
+      "step": 997
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.22365988909427e-05,
+      "loss": 17.7972,
+      "step": 998
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.23290203327172e-05,
+      "loss": 19.615,
+      "step": 999
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.242144177449169e-05,
+      "loss": 18.351,
+      "step": 1000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.251386321626618e-05,
+      "loss": 18.9204,
+      "step": 1001
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.260628465804067e-05,
+      "loss": 18.3205,
+      "step": 1002
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.269870609981516e-05,
+      "loss": 19.0011,
+      "step": 1003
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.279112754158965e-05,
+      "loss": 16.4721,
+      "step": 1004
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.288354898336414e-05,
+      "loss": 21.1151,
+      "step": 1005
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.297597042513864e-05,
+      "loss": 17.9361,
+      "step": 1006
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.306839186691313e-05,
+      "loss": 16.9099,
+      "step": 1007
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.316081330868762e-05,
+      "loss": 15.8302,
+      "step": 1008
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.325323475046212e-05,
+      "loss": 17.096,
+      "step": 1009
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.334565619223661e-05,
+      "loss": 17.2157,
+      "step": 1010
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.343807763401109e-05,
+      "loss": 17.2243,
+      "step": 1011
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.353049907578559e-05,
+      "loss": 17.3034,
+      "step": 1012
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.362292051756008e-05,
+      "loss": 18.7117,
+      "step": 1013
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.371534195933457e-05,
+      "loss": 17.4639,
+      "step": 1014
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.380776340110906e-05,
+      "loss": 17.4952,
+      "step": 1015
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.390018484288356e-05,
+      "loss": 18.3234,
+      "step": 1016
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.399260628465805e-05,
+      "loss": 16.2954,
+      "step": 1017
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.408502772643254e-05,
+      "loss": 16.5226,
+      "step": 1018
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.417744916820703e-05,
+      "loss": 16.3499,
+      "step": 1019
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.426987060998153e-05,
+      "loss": 15.9459,
+      "step": 1020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.4362292051756e-05,
+      "loss": 19.7914,
+      "step": 1021
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.44547134935305e-05,
+      "loss": 18.8314,
+      "step": 1022
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.4547134935305e-05,
+      "loss": 18.6312,
+      "step": 1023
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.463955637707949e-05,
+      "loss": 18.1163,
+      "step": 1024
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.473197781885397e-05,
+      "loss": 18.3732,
+      "step": 1025
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.482439926062848e-05,
+      "loss": 16.7421,
+      "step": 1026
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.491682070240297e-05,
+      "loss": 16.422,
+      "step": 1027
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.500924214417746e-05,
+      "loss": 17.7741,
+      "step": 1028
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.510166358595194e-05,
+      "loss": 17.1363,
+      "step": 1029
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.519408502772643e-05,
+      "loss": 17.8765,
+      "step": 1030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.528650646950092e-05,
+      "loss": 19.2891,
+      "step": 1031
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.537892791127541e-05,
+      "loss": 17.0526,
+      "step": 1032
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.547134935304991e-05,
+      "loss": 17.7398,
+      "step": 1033
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.55637707948244e-05,
+      "loss": 17.546,
+      "step": 1034
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.565619223659889e-05,
+      "loss": 17.5304,
+      "step": 1035
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.574861367837338e-05,
+      "loss": 19.5278,
+      "step": 1036
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.584103512014788e-05,
+      "loss": 16.6598,
+      "step": 1037
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.593345656192237e-05,
+      "loss": 18.1052,
+      "step": 1038
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.602587800369686e-05,
+      "loss": 18.4522,
+      "step": 1039
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.611829944547135e-05,
+      "loss": 17.1346,
+      "step": 1040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.621072088724584e-05,
+      "loss": 19.4718,
+      "step": 1041
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.630314232902033e-05,
+      "loss": 19.0216,
+      "step": 1042
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.639556377079483e-05,
+      "loss": 17.9273,
+      "step": 1043
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.648798521256932e-05,
+      "loss": 19.5327,
+      "step": 1044
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.658040665434381e-05,
+      "loss": 17.2417,
+      "step": 1045
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.66728280961183e-05,
+      "loss": 20.0021,
+      "step": 1046
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.67652495378928e-05,
+      "loss": 19.5922,
+      "step": 1047
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.685767097966729e-05,
+      "loss": 18.5739,
+      "step": 1048
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.695009242144177e-05,
+      "loss": 19.0506,
+      "step": 1049
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.704251386321627e-05,
+      "loss": 17.79,
+      "step": 1050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.713493530499076e-05,
+      "loss": 20.0115,
+      "step": 1051
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.722735674676525e-05,
+      "loss": 19.9614,
+      "step": 1052
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.731977818853975e-05,
+      "loss": 18.5813,
+      "step": 1053
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.741219963031424e-05,
+      "loss": 18.727,
+      "step": 1054
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.750462107208873e-05,
+      "loss": 19.547,
+      "step": 1055
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.759704251386322e-05,
+      "loss": 17.5274,
+      "step": 1056
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.768946395563772e-05,
+      "loss": 16.677,
+      "step": 1057
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.778188539741221e-05,
+      "loss": 18.3518,
+      "step": 1058
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.787430683918669e-05,
+      "loss": 17.0078,
+      "step": 1059
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.796672828096119e-05,
+      "loss": 17.0663,
+      "step": 1060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.805914972273568e-05,
+      "loss": 18.4223,
+      "step": 1061
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.815157116451017e-05,
+      "loss": 18.1152,
+      "step": 1062
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.824399260628466e-05,
+      "loss": 16.5873,
+      "step": 1063
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.833641404805916e-05,
+      "loss": 17.1161,
+      "step": 1064
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.842883548983365e-05,
+      "loss": 19.6622,
+      "step": 1065
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.852125693160814e-05,
+      "loss": 19.6828,
+      "step": 1066
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.861367837338264e-05,
+      "loss": 18.7116,
+      "step": 1067
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.870609981515712e-05,
+      "loss": 17.9451,
+      "step": 1068
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.87985212569316e-05,
+      "loss": 20.1147,
+      "step": 1069
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.889094269870611e-05,
+      "loss": 16.8102,
+      "step": 1070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89833641404806e-05,
+      "loss": 19.9852,
+      "step": 1071
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907578558225509e-05,
+      "loss": 17.2491,
+      "step": 1072
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916820702402958e-05,
+      "loss": 20.5246,
+      "step": 1073
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.926062846580408e-05,
+      "loss": 19.1928,
+      "step": 1074
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.935304990757857e-05,
+      "loss": 18.0543,
+      "step": 1075
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.944547134935306e-05,
+      "loss": 16.0091,
+      "step": 1076
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.953789279112755e-05,
+      "loss": 17.6004,
+      "step": 1077
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.963031423290203e-05,
+      "loss": 17.8893,
+      "step": 1078
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.972273567467652e-05,
+      "loss": 16.5574,
+      "step": 1079
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.981515711645101e-05,
+      "loss": 17.2425,
+      "step": 1080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.990757855822552e-05,
+      "loss": 18.7354,
+      "step": 1081
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001,
+      "loss": 16.1151,
+      "step": 1082
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999999979811832e-05,
+      "loss": 16.6416,
+      "step": 1083
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.99999991924733e-05,
+      "loss": 16.613,
+      "step": 1084
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999999818306491e-05,
+      "loss": 18.6851,
+      "step": 1085
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999999676989319e-05,
+      "loss": 19.2927,
+      "step": 1086
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999999495295813e-05,
+      "loss": 16.7445,
+      "step": 1087
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999999273225978e-05,
+      "loss": 19.2416,
+      "step": 1088
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.99999901077981e-05,
+      "loss": 17.1042,
+      "step": 1089
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999998707957316e-05,
+      "loss": 16.9166,
+      "step": 1090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999998364758496e-05,
+      "loss": 17.7362,
+      "step": 1091
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999997981183356e-05,
+      "loss": 17.0112,
+      "step": 1092
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999997557231894e-05,
+      "loss": 17.0645,
+      "step": 1093
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999997092904116e-05,
+      "loss": 17.7951,
+      "step": 1094
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999996588200028e-05,
+      "loss": 17.4936,
+      "step": 1095
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.99999604311963e-05,
+      "loss": 17.6917,
+      "step": 1096
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999995457662929e-05,
+      "loss": 17.1424,
+      "step": 1097
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999994831829931e-05,
+      "loss": 18.4615,
+      "step": 1098
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999994165620637e-05,
+      "loss": 16.0428,
+      "step": 1099
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999993459035055e-05,
+      "loss": 19.1281,
+      "step": 1100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.99999271207319e-05,
+      "loss": 19.6642,
+      "step": 1101
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999991924735048e-05,
+      "loss": 17.5356,
+      "step": 1102
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999991097020636e-05,
+      "loss": 15.2812,
+      "step": 1103
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999990228929961e-05,
+      "loss": 18.9522,
+      "step": 1104
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999989320463028e-05,
+      "loss": 17.9384,
+      "step": 1105
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999988371619847e-05,
+      "loss": 17.9909,
+      "step": 1106
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999987382400425e-05,
+      "loss": 18.4076,
+      "step": 1107
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999986352804767e-05,
+      "loss": 20.195,
+      "step": 1108
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999985282832885e-05,
+      "loss": 18.2947,
+      "step": 1109
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999984172484784e-05,
+      "loss": 16.8237,
+      "step": 1110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999983021760478e-05,
+      "loss": 18.46,
+      "step": 1111
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999981830659973e-05,
+      "loss": 19.5671,
+      "step": 1112
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999980599183278e-05,
+      "loss": 18.1329,
+      "step": 1113
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999979327330405e-05,
+      "loss": 16.7652,
+      "step": 1114
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999978015101363e-05,
+      "loss": 18.7556,
+      "step": 1115
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999976662496162e-05,
+      "loss": 17.7483,
+      "step": 1116
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999975269514817e-05,
+      "loss": 17.9467,
+      "step": 1117
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999973836157333e-05,
+      "loss": 18.5455,
+      "step": 1118
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999972362423725e-05,
+      "loss": 17.1202,
+      "step": 1119
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999970848314005e-05,
+      "loss": 20.0838,
+      "step": 1120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999969293828186e-05,
+      "loss": 16.693,
+      "step": 1121
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999967698966278e-05,
+      "loss": 17.7718,
+      "step": 1122
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999966063728295e-05,
+      "loss": 16.1417,
+      "step": 1123
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999964388114251e-05,
+      "loss": 17.6617,
+      "step": 1124
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.99996267212416e-05,
+      "loss": 16.9475,
+      "step": 1125
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999960915758034e-05,
+      "loss": 15.3515,
+      "step": 1126
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999959119015887e-05,
+      "loss": 16.2436,
+      "step": 1127
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999957281897735e-05,
+      "loss": 18.2292,
+      "step": 1128
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999955404403594e-05,
+      "loss": 16.7783,
+      "step": 1129
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999953486533475e-05,
+      "loss": 16.7641,
+      "step": 1130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999951528287396e-05,
+      "loss": 19.1909,
+      "step": 1131
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999949529665375e-05,
+      "loss": 17.7329,
+      "step": 1132
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999947490667425e-05,
+      "loss": 18.0334,
+      "step": 1133
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999945411293564e-05,
+      "loss": 15.9907,
+      "step": 1134
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999943291543808e-05,
+      "loss": 19.0897,
+      "step": 1135
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999941131418173e-05,
+      "loss": 18.3984,
+      "step": 1136
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999938930916679e-05,
+      "loss": 16.1539,
+      "step": 1137
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999936690039342e-05,
+      "loss": 19.058,
+      "step": 1138
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999934408786183e-05,
+      "loss": 18.6816,
+      "step": 1139
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999932087157215e-05,
+      "loss": 16.7049,
+      "step": 1140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.999929725152461e-05,
+      "loss": 16.2489,
+      "step": 1141
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999927322771938e-05,
+      "loss": 16.9994,
+      "step": 1142
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999924880015668e-05,
+      "loss": 20.1412,
+      "step": 1143
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999922396883667e-05,
+      "loss": 18.9634,
+      "step": 1144
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999919873375958e-05,
+      "loss": 18.0686,
+      "step": 1145
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999917309492562e-05,
+      "loss": 16.6736,
+      "step": 1146
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999914705233497e-05,
+      "loss": 15.863,
+      "step": 1147
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999912060598785e-05,
+      "loss": 18.0268,
+      "step": 1148
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999909375588446e-05,
+      "loss": 16.269,
+      "step": 1149
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999906650202505e-05,
+      "loss": 18.5387,
+      "step": 1150
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999903884440981e-05,
+      "loss": 18.2787,
+      "step": 1151
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999901078303899e-05,
+      "loss": 16.8413,
+      "step": 1152
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.99989823179128e-05,
+      "loss": 17.08,
+      "step": 1153
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999895344903148e-05,
+      "loss": 15.8339,
+      "step": 1154
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999892417639523e-05,
+      "loss": 17.9789,
+      "step": 1155
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999889450000433e-05,
+      "loss": 18.1745,
+      "step": 1156
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999886441985899e-05,
+      "loss": 17.1007,
+      "step": 1157
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999883393595947e-05,
+      "loss": 17.2272,
+      "step": 1158
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999880304830602e-05,
+      "loss": 19.8662,
+      "step": 1159
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999877175689887e-05,
+      "loss": 17.5368,
+      "step": 1160
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999874006173828e-05,
+      "loss": 17.5884,
+      "step": 1161
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999870796282451e-05,
+      "loss": 19.5856,
+      "step": 1162
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999867546015782e-05,
+      "loss": 16.3115,
+      "step": 1163
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999864255373848e-05,
+      "loss": 17.3903,
+      "step": 1164
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999860924356672e-05,
+      "loss": 17.9746,
+      "step": 1165
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999857552964285e-05,
+      "loss": 17.9088,
+      "step": 1166
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999854141196714e-05,
+      "loss": 16.9663,
+      "step": 1167
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999850689053982e-05,
+      "loss": 15.8265,
+      "step": 1168
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999847196536123e-05,
+      "loss": 16.156,
+      "step": 1169
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999843663643161e-05,
+      "loss": 17.5493,
+      "step": 1170
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999840090375126e-05,
+      "loss": 18.4059,
+      "step": 1171
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999836476732048e-05,
+      "loss": 17.2559,
+      "step": 1172
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999832822713953e-05,
+      "loss": 18.1325,
+      "step": 1173
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999829128320874e-05,
+      "loss": 14.6887,
+      "step": 1174
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999825393552838e-05,
+      "loss": 17.0418,
+      "step": 1175
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999821618409877e-05,
+      "loss": 18.5891,
+      "step": 1176
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.99981780289202e-05,
+      "loss": 16.2717,
+      "step": 1177
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.9998139469993e-05,
+      "loss": 18.4196,
+      "step": 1178
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999810050731746e-05,
+      "loss": 16.0231,
+      "step": 1179
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.99980611408939e-05,
+      "loss": 17.5884,
+      "step": 1180
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999802137072266e-05,
+      "loss": 17.4614,
+      "step": 1181
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999798119680402e-05,
+      "loss": 17.9079,
+      "step": 1182
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999794061913834e-05,
+      "loss": 17.9886,
+      "step": 1183
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999789963772593e-05,
+      "loss": 18.6554,
+      "step": 1184
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999785825256712e-05,
+      "loss": 16.5567,
+      "step": 1185
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999781646366225e-05,
+      "loss": 16.7344,
+      "step": 1186
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999777427101167e-05,
+      "loss": 17.5103,
+      "step": 1187
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999773167461568e-05,
+      "loss": 15.9357,
+      "step": 1188
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999768867447468e-05,
+      "loss": 18.6588,
+      "step": 1189
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999764527058897e-05,
+      "loss": 17.814,
+      "step": 1190
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999760146295893e-05,
+      "loss": 18.4814,
+      "step": 1191
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.99975572515849e-05,
+      "loss": 18.7757,
+      "step": 1192
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999751263646724e-05,
+      "loss": 15.6231,
+      "step": 1193
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999746761760631e-05,
+      "loss": 17.5581,
+      "step": 1194
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999742219500247e-05,
+      "loss": 18.3567,
+      "step": 1195
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999737636865609e-05,
+      "loss": 16.3479,
+      "step": 1196
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999733013856754e-05,
+      "loss": 15.8529,
+      "step": 1197
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999728350473721e-05,
+      "loss": 14.4003,
+      "step": 1198
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999723646716544e-05,
+      "loss": 17.234,
+      "step": 1199
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999718902585264e-05,
+      "loss": 18.3516,
+      "step": 1200
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999714118079917e-05,
+      "loss": 16.3276,
+      "step": 1201
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999709293200546e-05,
+      "loss": 19.0791,
+      "step": 1202
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999704427947182e-05,
+      "loss": 17.7069,
+      "step": 1203
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999699522319874e-05,
+      "loss": 18.8735,
+      "step": 1204
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999694576318654e-05,
+      "loss": 17.4308,
+      "step": 1205
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999689589943566e-05,
+      "loss": 17.185,
+      "step": 1206
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999684563194649e-05,
+      "loss": 18.1052,
+      "step": 1207
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999679496071944e-05,
+      "loss": 16.7205,
+      "step": 1208
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999674388575491e-05,
+      "loss": 16.7494,
+      "step": 1209
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999669240705332e-05,
+      "loss": 18.5328,
+      "step": 1210
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999664052461509e-05,
+      "loss": 17.7166,
+      "step": 1211
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999658823844062e-05,
+      "loss": 17.0417,
+      "step": 1212
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999653554853035e-05,
+      "loss": 19.0853,
+      "step": 1213
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.99964824548847e-05,
+      "loss": 17.6405,
+      "step": 1214
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.99964289575041e-05,
+      "loss": 17.4694,
+      "step": 1215
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999637505638898e-05,
+      "loss": 17.3932,
+      "step": 1216
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999632075153977e-05,
+      "loss": 14.9172,
+      "step": 1217
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999626604295693e-05,
+      "loss": 17.7602,
+      "step": 1218
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999621093064088e-05,
+      "loss": 17.9841,
+      "step": 1219
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999615541459207e-05,
+      "loss": 17.0323,
+      "step": 1220
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999609949481095e-05,
+      "loss": 17.7313,
+      "step": 1221
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999604317129797e-05,
+      "loss": 18.1472,
+      "step": 1222
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999598644405359e-05,
+      "loss": 17.3105,
+      "step": 1223
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999592931307827e-05,
+      "loss": 18.4173,
+      "step": 1224
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999587177837245e-05,
+      "loss": 17.2256,
+      "step": 1225
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999581383993663e-05,
+      "loss": 18.418,
+      "step": 1226
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999575549777124e-05,
+      "loss": 16.2123,
+      "step": 1227
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999569675187678e-05,
+      "loss": 19.0155,
+      "step": 1228
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999563760225372e-05,
+      "loss": 18.1758,
+      "step": 1229
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999557804890252e-05,
+      "loss": 16.8194,
+      "step": 1230
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999551809182367e-05,
+      "loss": 19.2069,
+      "step": 1231
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999545773101766e-05,
+      "loss": 17.2779,
+      "step": 1232
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999539696648497e-05,
+      "loss": 18.5662,
+      "step": 1233
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.99953357982261e-05,
+      "loss": 18.0267,
+      "step": 1234
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999527422624155e-05,
+      "loss": 17.2475,
+      "step": 1235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999521225053179e-05,
+      "loss": 17.7811,
+      "step": 1236
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999514987109735e-05,
+      "loss": 17.6795,
+      "step": 1237
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.99950870879387e-05,
+      "loss": 16.2717,
+      "step": 1238
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999502390105638e-05,
+      "loss": 18.2378,
+      "step": 1239
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999496031045087e-05,
+      "loss": 19.1898,
+      "step": 1240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999489631612273e-05,
+      "loss": 16.8908,
+      "step": 1241
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999483191807244e-05,
+      "loss": 18.5135,
+      "step": 1242
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999476711630051e-05,
+      "loss": 16.6189,
+      "step": 1243
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.99947019108075e-05,
+      "loss": 17.8355,
+      "step": 1244
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999463630159391e-05,
+      "loss": 19.2753,
+      "step": 1245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999457028866028e-05,
+      "loss": 18.6581,
+      "step": 1246
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999450387200714e-05,
+      "loss": 15.2966,
+      "step": 1247
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999443705163503e-05,
+      "loss": 17.0196,
+      "step": 1248
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999436982754448e-05,
+      "loss": 16.9061,
+      "step": 1249
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999430219973604e-05,
+      "loss": 17.3797,
+      "step": 1250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999423416821026e-05,
+      "loss": 18.0759,
+      "step": 1251
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999416573296767e-05,
+      "loss": 17.1538,
+      "step": 1252
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999409689400885e-05,
+      "loss": 15.7275,
+      "step": 1253
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999402765133434e-05,
+      "loss": 17.4918,
+      "step": 1254
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.99939580049447e-05,
+      "loss": 17.2145,
+      "step": 1255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.99938879548405e-05,
+      "loss": 17.2612,
+      "step": 1256
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.99938175010223e-05,
+      "loss": 19.1962,
+      "step": 1257
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999374664349068e-05,
+      "loss": 18.4513,
+      "step": 1258
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999367538224617e-05,
+      "loss": 17.3925,
+      "step": 1259
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.99936037172894e-05,
+      "loss": 19.271,
+      "step": 1260
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.999353164862094e-05,
+      "loss": 19.0329,
+      "step": 1261
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999345917624132e-05,
+      "loss": 17.3844,
+      "step": 1262
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999338630015119e-05,
+      "loss": 20.2166,
+      "step": 1263
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999331302035111e-05,
+      "loss": 17.9928,
+      "step": 1264
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999323933684166e-05,
+      "loss": 18.0423,
+      "step": 1265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999316524962345e-05,
+      "loss": 19.6966,
+      "step": 1266
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999309075869708e-05,
+      "loss": 17.2661,
+      "step": 1267
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999301586406316e-05,
+      "loss": 18.7408,
+      "step": 1268
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999294056572226e-05,
+      "loss": 18.0739,
+      "step": 1269
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999286486367503e-05,
+      "loss": 16.7544,
+      "step": 1270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999278875792206e-05,
+      "loss": 17.4051,
+      "step": 1271
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999271224846396e-05,
+      "loss": 17.2488,
+      "step": 1272
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999263533530135e-05,
+      "loss": 16.8554,
+      "step": 1273
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999255801843486e-05,
+      "loss": 18.1299,
+      "step": 1274
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999248029786512e-05,
+      "loss": 16.249,
+      "step": 1275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999240217359274e-05,
+      "loss": 17.037,
+      "step": 1276
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999232364561835e-05,
+      "loss": 17.9456,
+      "step": 1277
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.99922447139426e-05,
+      "loss": 23.1127,
+      "step": 1278
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999216537856612e-05,
+      "loss": 19.1514,
+      "step": 1279
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999208563948956e-05,
+      "loss": 16.8461,
+      "step": 1280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999200549671355e-05,
+      "loss": 17.1358,
+      "step": 1281
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999192495023873e-05,
+      "loss": 17.9036,
+      "step": 1282
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999184400006577e-05,
+      "loss": 17.3549,
+      "step": 1283
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999176264619531e-05,
+      "loss": 18.1054,
+      "step": 1284
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999168088862803e-05,
+      "loss": 16.0531,
+      "step": 1285
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999159872736456e-05,
+      "loss": 17.9164,
+      "step": 1286
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999151616240557e-05,
+      "loss": 17.3254,
+      "step": 1287
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999143319375174e-05,
+      "loss": 18.0932,
+      "step": 1288
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999134982140375e-05,
+      "loss": 16.9479,
+      "step": 1289
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999126604536223e-05,
+      "loss": 16.843,
+      "step": 1290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.99911818656279e-05,
+      "loss": 16.4307,
+      "step": 1291
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999109728220142e-05,
+      "loss": 18.8322,
+      "step": 1292
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999101229508346e-05,
+      "loss": 17.0678,
+      "step": 1293
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999092690427474e-05,
+      "loss": 14.6777,
+      "step": 1294
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999084110977593e-05,
+      "loss": 19.8746,
+      "step": 1295
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999075491158772e-05,
+      "loss": 20.078,
+      "step": 1296
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.99906683097108e-05,
+      "loss": 19.1163,
+      "step": 1297
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.99905813041459e-05,
+      "loss": 19.0595,
+      "step": 1298
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.99904938948937e-05,
+      "loss": 17.9403,
+      "step": 1299
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.99904060819549e-05,
+      "loss": 19.0498,
+      "step": 1300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999031786533022e-05,
+      "loss": 15.7938,
+      "step": 1301
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999022924502037e-05,
+      "loss": 17.534,
+      "step": 1302
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999014022102606e-05,
+      "loss": 18.2206,
+      "step": 1303
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.999005079334803e-05,
+      "loss": 18.544,
+      "step": 1304
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998996096198697e-05,
+      "loss": 18.6722,
+      "step": 1305
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998987072694363e-05,
+      "loss": 18.7255,
+      "step": 1306
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998978008821872e-05,
+      "loss": 17.2336,
+      "step": 1307
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.9989689045813e-05,
+      "loss": 17.893,
+      "step": 1308
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998959759972717e-05,
+      "loss": 16.5591,
+      "step": 1309
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998950574996199e-05,
+      "loss": 17.4984,
+      "step": 1310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.99894134965182e-05,
+      "loss": 17.5527,
+      "step": 1311
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998932083939656e-05,
+      "loss": 17.9683,
+      "step": 1312
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998922777859778e-05,
+      "loss": 17.5353,
+      "step": 1313
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998913431412264e-05,
+      "loss": 20.8,
+      "step": 1314
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998904044597188e-05,
+      "loss": 16.6334,
+      "step": 1315
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998894617414628e-05,
+      "loss": 18.3546,
+      "step": 1316
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998885149864658e-05,
+      "loss": 16.6275,
+      "step": 1317
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998875641947354e-05,
+      "loss": 17.0088,
+      "step": 1318
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998866093662796e-05,
+      "loss": 16.69,
+      "step": 1319
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998856505011059e-05,
+      "loss": 16.1314,
+      "step": 1320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998846875992219e-05,
+      "loss": 17.6137,
+      "step": 1321
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998837206606354e-05,
+      "loss": 18.4339,
+      "step": 1322
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998827496853544e-05,
+      "loss": 16.9348,
+      "step": 1323
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998817746733868e-05,
+      "loss": 17.0761,
+      "step": 1324
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998807956247402e-05,
+      "loss": 17.4387,
+      "step": 1325
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998798125394228e-05,
+      "loss": 19.0782,
+      "step": 1326
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998788254174424e-05,
+      "loss": 17.965,
+      "step": 1327
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998778342588069e-05,
+      "loss": 18.123,
+      "step": 1328
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998768390635242e-05,
+      "loss": 17.0586,
+      "step": 1329
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998758398316027e-05,
+      "loss": 16.6446,
+      "step": 1330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998748365630502e-05,
+      "loss": 19.1503,
+      "step": 1331
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998738292578749e-05,
+      "loss": 18.8551,
+      "step": 1332
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998728179160848e-05,
+      "loss": 17.5034,
+      "step": 1333
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998718025376884e-05,
+      "loss": 17.5502,
+      "step": 1334
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998707831226934e-05,
+      "loss": 16.2033,
+      "step": 1335
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998697596711084e-05,
+      "loss": 19.4234,
+      "step": 1336
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998687321829416e-05,
+      "loss": 16.3601,
+      "step": 1337
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998677006582011e-05,
+      "loss": 15.6891,
+      "step": 1338
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998666650968955e-05,
+      "loss": 18.0254,
+      "step": 1339
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998656254990332e-05,
+      "loss": 17.9128,
+      "step": 1340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998645818646224e-05,
+      "loss": 17.1626,
+      "step": 1341
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998635341936714e-05,
+      "loss": 18.4926,
+      "step": 1342
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.99862482486189e-05,
+      "loss": 18.8206,
+      "step": 1343
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998614267421835e-05,
+      "loss": 16.2776,
+      "step": 1344
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998603669616634e-05,
+      "loss": 16.1801,
+      "step": 1345
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998593031446373e-05,
+      "loss": 16.2899,
+      "step": 1346
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998582352911139e-05,
+      "loss": 17.3718,
+      "step": 1347
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998571634011015e-05,
+      "loss": 17.7409,
+      "step": 1348
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998560874746094e-05,
+      "loss": 18.0028,
+      "step": 1349
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998550075116455e-05,
+      "loss": 17.4351,
+      "step": 1350
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.99853923512219e-05,
+      "loss": 18.3483,
+      "step": 1351
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998528354763387e-05,
+      "loss": 17.6633,
+      "step": 1352
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998517434040131e-05,
+      "loss": 17.6149,
+      "step": 1353
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998506472952511e-05,
+      "loss": 17.0618,
+      "step": 1354
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998495471500617e-05,
+      "loss": 18.4084,
+      "step": 1355
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998484429684537e-05,
+      "loss": 17.7616,
+      "step": 1356
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998473347504358e-05,
+      "loss": 19.204,
+      "step": 1357
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998462224960175e-05,
+      "loss": 17.099,
+      "step": 1358
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998451062052072e-05,
+      "loss": 16.3182,
+      "step": 1359
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998439858780142e-05,
+      "loss": 19.9097,
+      "step": 1360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998428615144475e-05,
+      "loss": 18.419,
+      "step": 1361
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998417331145161e-05,
+      "loss": 18.7555,
+      "step": 1362
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998406006782293e-05,
+      "loss": 17.7869,
+      "step": 1363
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.99839464205596e-05,
+      "loss": 18.3104,
+      "step": 1364
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998383236966256e-05,
+      "loss": 18.1685,
+      "step": 1365
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998371791513272e-05,
+      "loss": 18.7297,
+      "step": 1366
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998360305697101e-05,
+      "loss": 17.3875,
+      "step": 1367
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998348779517835e-05,
+      "loss": 16.7808,
+      "step": 1368
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998337212975568e-05,
+      "loss": 17.4915,
+      "step": 1369
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998325606070391e-05,
+      "loss": 18.267,
+      "step": 1370
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998313958802401e-05,
+      "loss": 18.4768,
+      "step": 1371
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998302271171691e-05,
+      "loss": 20.2388,
+      "step": 1372
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998290543178355e-05,
+      "loss": 17.373,
+      "step": 1373
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998278774822487e-05,
+      "loss": 16.5209,
+      "step": 1374
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998266966104183e-05,
+      "loss": 17.942,
+      "step": 1375
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998255117023538e-05,
+      "loss": 17.9644,
+      "step": 1376
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998243227580648e-05,
+      "loss": 18.1579,
+      "step": 1377
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998231297775609e-05,
+      "loss": 18.1129,
+      "step": 1378
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998219327608518e-05,
+      "loss": 17.0672,
+      "step": 1379
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998207317079469e-05,
+      "loss": 19.0978,
+      "step": 1380
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998195266188562e-05,
+      "loss": 16.0283,
+      "step": 1381
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.998183174935893e-05,
+      "loss": 18.6603,
+      "step": 1382
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.998171043321558e-05,
+      "loss": 18.2169,
+      "step": 1383
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.998158871345658e-05,
+      "loss": 17.566,
+      "step": 1384
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.998146659008291e-05,
+      "loss": 17.7085,
+      "step": 1385
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.998134406309554e-05,
+      "loss": 18.2622,
+      "step": 1386
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.998122113249546e-05,
+      "loss": 17.5868,
+      "step": 1387
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.998109779828368e-05,
+      "loss": 18.9292,
+      "step": 1388
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.998097406046116e-05,
+      "loss": 17.2718,
+      "step": 1389
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.998084991902894e-05,
+      "loss": 16.1467,
+      "step": 1390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.998072537398801e-05,
+      "loss": 19.396,
+      "step": 1391
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.998060042533936e-05,
+      "loss": 18.2235,
+      "step": 1392
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.998047507308404e-05,
+      "loss": 16.6192,
+      "step": 1393
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.9980349317223e-05,
+      "loss": 16.8762,
+      "step": 1394
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.99802231577573e-05,
+      "loss": 16.8397,
+      "step": 1395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.998009659468794e-05,
+      "loss": 18.1367,
+      "step": 1396
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997996962801595e-05,
+      "loss": 18.7757,
+      "step": 1397
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997984225774238e-05,
+      "loss": 17.7908,
+      "step": 1398
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.99797144838682e-05,
+      "loss": 18.1388,
+      "step": 1399
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997958630639449e-05,
+      "loss": 18.3307,
+      "step": 1400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997945772532227e-05,
+      "loss": 18.1314,
+      "step": 1401
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997932874065258e-05,
+      "loss": 19.4409,
+      "step": 1402
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997919935238646e-05,
+      "loss": 17.2037,
+      "step": 1403
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997906956052494e-05,
+      "loss": 16.4672,
+      "step": 1404
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.99789393650691e-05,
+      "loss": 18.1751,
+      "step": 1405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997880876601997e-05,
+      "loss": 17.4237,
+      "step": 1406
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997867776337862e-05,
+      "loss": 17.8114,
+      "step": 1407
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997854635714607e-05,
+      "loss": 17.6862,
+      "step": 1408
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997841454732341e-05,
+      "loss": 15.6447,
+      "step": 1409
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997828233391173e-05,
+      "loss": 16.686,
+      "step": 1410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997814971691206e-05,
+      "loss": 17.8978,
+      "step": 1411
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997801669632547e-05,
+      "loss": 17.6376,
+      "step": 1412
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997788327215305e-05,
+      "loss": 19.4904,
+      "step": 1413
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997774944439587e-05,
+      "loss": 19.5517,
+      "step": 1414
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997761521305501e-05,
+      "loss": 17.5634,
+      "step": 1415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997748057813157e-05,
+      "loss": 17.8348,
+      "step": 1416
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997734553962662e-05,
+      "loss": 16.9461,
+      "step": 1417
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997721009754125e-05,
+      "loss": 17.6314,
+      "step": 1418
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997707425187657e-05,
+      "loss": 18.2224,
+      "step": 1419
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997693800263366e-05,
+      "loss": 17.6926,
+      "step": 1420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997680134981362e-05,
+      "loss": 18.0254,
+      "step": 1421
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997666429341755e-05,
+      "loss": 16.0795,
+      "step": 1422
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997652683344658e-05,
+      "loss": 17.7658,
+      "step": 1423
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.99763889699018e-05,
+      "loss": 16.8742,
+      "step": 1424
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997625070278433e-05,
+      "loss": 15.7762,
+      "step": 1425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997611203209529e-05,
+      "loss": 18.8038,
+      "step": 1426
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.99759729578358e-05,
+      "loss": 16.4403,
+      "step": 1427
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997583348000698e-05,
+      "loss": 20.6615,
+      "step": 1428
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997569359860995e-05,
+      "loss": 17.1281,
+      "step": 1429
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997555331364584e-05,
+      "loss": 20.0932,
+      "step": 1430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997541262511579e-05,
+      "loss": 19.3113,
+      "step": 1431
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997527153302093e-05,
+      "loss": 18.8098,
+      "step": 1432
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.99751300373624e-05,
+      "loss": 17.1202,
+      "step": 1433
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997498813814137e-05,
+      "loss": 17.0164,
+      "step": 1434
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997484583535895e-05,
+      "loss": 17.0173,
+      "step": 1435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997470312901629e-05,
+      "loss": 17.6352,
+      "step": 1436
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997456001911457e-05,
+      "loss": 18.836,
+      "step": 1437
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.99744165056549e-05,
+      "loss": 18.7769,
+      "step": 1438
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997427258863849e-05,
+      "loss": 19.4202,
+      "step": 1439
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997412826806647e-05,
+      "loss": 18.3154,
+      "step": 1440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997398354394002e-05,
+      "loss": 18.4537,
+      "step": 1441
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.99738384162603e-05,
+      "loss": 18.1061,
+      "step": 1442
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997369288502849e-05,
+      "loss": 16.2622,
+      "step": 1443
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997354695024577e-05,
+      "loss": 16.8989,
+      "step": 1444
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.99734006119133e-05,
+      "loss": 19.1215,
+      "step": 1445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997325387003227e-05,
+      "loss": 16.6289,
+      "step": 1446
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997310672460386e-05,
+      "loss": 18.4142,
+      "step": 1447
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997295917562927e-05,
+      "loss": 19.2263,
+      "step": 1448
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997281122310968e-05,
+      "loss": 15.0355,
+      "step": 1449
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997266286704631e-05,
+      "loss": 19.0349,
+      "step": 1450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997251410744033e-05,
+      "loss": 17.5298,
+      "step": 1451
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997236494429294e-05,
+      "loss": 16.5046,
+      "step": 1452
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997221537760536e-05,
+      "loss": 15.4263,
+      "step": 1453
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997206540737878e-05,
+      "loss": 15.736,
+      "step": 1454
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997191503361445e-05,
+      "loss": 17.1897,
+      "step": 1455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997176425631354e-05,
+      "loss": 17.949,
+      "step": 1456
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997161307547729e-05,
+      "loss": 17.7187,
+      "step": 1457
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997146149110692e-05,
+      "loss": 15.746,
+      "step": 1458
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997130950320363e-05,
+      "loss": 16.1268,
+      "step": 1459
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.99711571117687e-05,
+      "loss": 18.6712,
+      "step": 1460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997100431680331e-05,
+      "loss": 18.1039,
+      "step": 1461
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.99708511183087e-05,
+      "loss": 19.2248,
+      "step": 1462
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997069751628614e-05,
+      "loss": 17.1199,
+      "step": 1463
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997054351073683e-05,
+      "loss": 16.4614,
+      "step": 1464
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997038910166206e-05,
+      "loss": 18.0068,
+      "step": 1465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.997023428906302e-05,
+      "loss": 16.408,
+      "step": 1466
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.9970079072941e-05,
+      "loss": 18.1616,
+      "step": 1467
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996992345329726e-05,
+      "loss": 16.7392,
+      "step": 1468
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996976743013302e-05,
+      "loss": 19.3213,
+      "step": 1469
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996961100344957e-05,
+      "loss": 16.9252,
+      "step": 1470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996945417324817e-05,
+      "loss": 15.4427,
+      "step": 1471
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996929693953007e-05,
+      "loss": 15.0606,
+      "step": 1472
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996913930229655e-05,
+      "loss": 16.8409,
+      "step": 1473
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996898126154888e-05,
+      "loss": 16.3615,
+      "step": 1474
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996882281728835e-05,
+      "loss": 18.9531,
+      "step": 1475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996866396951621e-05,
+      "loss": 16.1723,
+      "step": 1476
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996850471823376e-05,
+      "loss": 17.4096,
+      "step": 1477
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.99683450634423e-05,
+      "loss": 18.4603,
+      "step": 1478
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996818500514311e-05,
+      "loss": 16.5407,
+      "step": 1479
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996802454333747e-05,
+      "loss": 16.6651,
+      "step": 1480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996786367802668e-05,
+      "loss": 15.7732,
+      "step": 1481
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996770240921205e-05,
+      "loss": 18.8319,
+      "step": 1482
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996754073689488e-05,
+      "loss": 19.3013,
+      "step": 1483
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996737866107647e-05,
+      "loss": 18.3065,
+      "step": 1484
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996721618175812e-05,
+      "loss": 16.1893,
+      "step": 1485
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996705329894116e-05,
+      "loss": 16.2089,
+      "step": 1486
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.99668900126269e-05,
+      "loss": 17.1361,
+      "step": 1487
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996672632281664e-05,
+      "loss": 18.1316,
+      "step": 1488
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996656222951172e-05,
+      "loss": 18.2775,
+      "step": 1489
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996639773271347e-05,
+      "loss": 16.8298,
+      "step": 1490
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996623283242323e-05,
+      "loss": 17.4413,
+      "step": 1491
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996606752864229e-05,
+      "loss": 18.1851,
+      "step": 1492
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996590182137202e-05,
+      "loss": 17.8892,
+      "step": 1493
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996573571061374e-05,
+      "loss": 18.8688,
+      "step": 1494
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996556919636879e-05,
+      "loss": 19.2995,
+      "step": 1495
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996540227863854e-05,
+      "loss": 15.8158,
+      "step": 1496
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996523495742431e-05,
+      "loss": 18.7736,
+      "step": 1497
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996506723272745e-05,
+      "loss": 17.6402,
+      "step": 1498
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996489910454934e-05,
+      "loss": 17.6841,
+      "step": 1499
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996473057289132e-05,
+      "loss": 18.5193,
+      "step": 1500
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.996456163775474e-05,
+      "loss": 16.0666,
+      "step": 1501
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.9964392299141e-05,
+      "loss": 16.0367,
+      "step": 1502
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996422255705145e-05,
+      "loss": 19.4391,
+      "step": 1503
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996405241148743e-05,
+      "loss": 14.7292,
+      "step": 1504
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996388186245036e-05,
+      "loss": 16.7072,
+      "step": 1505
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996371090994158e-05,
+      "loss": 15.7127,
+      "step": 1506
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996353955396251e-05,
+      "loss": 17.339,
+      "step": 1507
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99633677945145e-05,
+      "loss": 18.9701,
+      "step": 1508
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996319563159895e-05,
+      "loss": 16.5342,
+      "step": 1509
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996302306521725e-05,
+      "loss": 16.7836,
+      "step": 1510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996285009537078e-05,
+      "loss": 17.6547,
+      "step": 1511
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996267672206096e-05,
+      "loss": 16.9368,
+      "step": 1512
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996250294528919e-05,
+      "loss": 16.7761,
+      "step": 1513
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996232876505685e-05,
+      "loss": 17.6947,
+      "step": 1514
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996215418136536e-05,
+      "loss": 19.1487,
+      "step": 1515
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996197919421614e-05,
+      "loss": 17.041,
+      "step": 1516
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996180380361057e-05,
+      "loss": 16.5084,
+      "step": 1517
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996162800955011e-05,
+      "loss": 18.5644,
+      "step": 1518
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996145181203615e-05,
+      "loss": 16.5986,
+      "step": 1519
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996127521107012e-05,
+      "loss": 16.655,
+      "step": 1520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996109820665345e-05,
+      "loss": 18.1265,
+      "step": 1521
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996092079878756e-05,
+      "loss": 17.9124,
+      "step": 1522
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99607429874739e-05,
+      "loss": 18.0402,
+      "step": 1523
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996056477271389e-05,
+      "loss": 17.7837,
+      "step": 1524
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996038615450897e-05,
+      "loss": 17.1107,
+      "step": 1525
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996020713286059e-05,
+      "loss": 17.5418,
+      "step": 1526
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996002770777019e-05,
+      "loss": 18.1157,
+      "step": 1527
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995984787923923e-05,
+      "loss": 17.0063,
+      "step": 1528
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995966764726915e-05,
+      "loss": 16.3738,
+      "step": 1529
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995948701186139e-05,
+      "loss": 16.0727,
+      "step": 1530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995930597301744e-05,
+      "loss": 18.4653,
+      "step": 1531
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995912453073874e-05,
+      "loss": 18.6491,
+      "step": 1532
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99589426850268e-05,
+      "loss": 21.6061,
+      "step": 1533
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995876043588301e-05,
+      "loss": 17.3238,
+      "step": 1534
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99585777833089e-05,
+      "loss": 16.0573,
+      "step": 1535
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995839472730593e-05,
+      "loss": 16.9589,
+      "step": 1536
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995821126787558e-05,
+      "loss": 18.0834,
+      "step": 1537
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995802740501933e-05,
+      "loss": 18.0201,
+      "step": 1538
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995784313873866e-05,
+      "loss": 17.1732,
+      "step": 1539
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995765846903507e-05,
+      "loss": 16.2126,
+      "step": 1540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995747339591004e-05,
+      "loss": 18.4542,
+      "step": 1541
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995728791936504e-05,
+      "loss": 16.288,
+      "step": 1542
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995710203940163e-05,
+      "loss": 18.4991,
+      "step": 1543
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995691575602127e-05,
+      "loss": 18.8299,
+      "step": 1544
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995672906922546e-05,
+      "loss": 18.978,
+      "step": 1545
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995654197901573e-05,
+      "loss": 19.1867,
+      "step": 1546
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995635448539358e-05,
+      "loss": 17.6205,
+      "step": 1547
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995616658836051e-05,
+      "loss": 16.2594,
+      "step": 1548
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995597828791807e-05,
+      "loss": 16.1316,
+      "step": 1549
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995578958406775e-05,
+      "loss": 18.1068,
+      "step": 1550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995560047681107e-05,
+      "loss": 18.6763,
+      "step": 1551
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99554109661496e-05,
+      "loss": 17.2533,
+      "step": 1552
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995522105208483e-05,
+      "loss": 18.5597,
+      "step": 1553
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995503073461829e-05,
+      "loss": 16.9917,
+      "step": 1554
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995484001375155e-05,
+      "loss": 16.5626,
+      "step": 1555
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995464888948614e-05,
+      "loss": 18.9472,
+      "step": 1556
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995445736182358e-05,
+      "loss": 16.8183,
+      "step": 1557
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995426543076545e-05,
+      "loss": 16.8528,
+      "step": 1558
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995407309631326e-05,
+      "loss": 16.1174,
+      "step": 1559
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99538803584686e-05,
+      "loss": 17.3231,
+      "step": 1560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995368721723301e-05,
+      "loss": 18.2708,
+      "step": 1561
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995349367260807e-05,
+      "loss": 16.6888,
+      "step": 1562
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995329972459529e-05,
+      "loss": 16.0542,
+      "step": 1563
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99531053731963e-05,
+      "loss": 17.41,
+      "step": 1564
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995291061841263e-05,
+      "loss": 17.4713,
+      "step": 1565
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995271546024587e-05,
+      "loss": 18.0443,
+      "step": 1566
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995251989869757e-05,
+      "loss": 17.4131,
+      "step": 1567
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995232393376934e-05,
+      "loss": 17.5124,
+      "step": 1568
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995212756546274e-05,
+      "loss": 17.6206,
+      "step": 1569
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99519307937794e-05,
+      "loss": 19.1289,
+      "step": 1570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995173361872086e-05,
+      "loss": 15.9466,
+      "step": 1571
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99515360402887e-05,
+      "loss": 18.3219,
+      "step": 1572
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995133805848457e-05,
+      "loss": 16.1484,
+      "step": 1573
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995113967331005e-05,
+      "loss": 16.155,
+      "step": 1574
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995094088476672e-05,
+      "loss": 18.0454,
+      "step": 1575
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99507416928562e-05,
+      "loss": 17.7757,
+      "step": 1576
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99505420975801e-05,
+      "loss": 18.7606,
+      "step": 1577
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995034209894003e-05,
+      "loss": 16.8777,
+      "step": 1578
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.995014169693761e-05,
+      "loss": 18.3428,
+      "step": 1579
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994994089157444e-05,
+      "loss": 15.6346,
+      "step": 1580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994973968285216e-05,
+      "loss": 19.0558,
+      "step": 1581
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994953807077239e-05,
+      "loss": 18.1914,
+      "step": 1582
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994933605533676e-05,
+      "loss": 17.7959,
+      "step": 1583
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99491336365469e-05,
+      "loss": 18.4641,
+      "step": 1584
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994893081440444e-05,
+      "loss": 17.1106,
+      "step": 1585
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994872758891104e-05,
+      "loss": 18.0863,
+      "step": 1586
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994852396006828e-05,
+      "loss": 16.1938,
+      "step": 1587
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994831992787788e-05,
+      "loss": 16.903,
+      "step": 1588
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994811549234144e-05,
+      "loss": 17.3814,
+      "step": 1589
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994791065346064e-05,
+      "loss": 16.8232,
+      "step": 1590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994770541123711e-05,
+      "loss": 18.2008,
+      "step": 1591
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994749976567253e-05,
+      "loss": 17.9391,
+      "step": 1592
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994729371676851e-05,
+      "loss": 16.8156,
+      "step": 1593
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994708726452679e-05,
+      "loss": 18.3923,
+      "step": 1594
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994688040894899e-05,
+      "loss": 17.5815,
+      "step": 1595
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994667315003678e-05,
+      "loss": 17.3599,
+      "step": 1596
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994646548779185e-05,
+      "loss": 18.0164,
+      "step": 1597
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994625742221586e-05,
+      "loss": 16.0375,
+      "step": 1598
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99460489533105e-05,
+      "loss": 17.4067,
+      "step": 1599
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994584008107746e-05,
+      "loss": 17.6505,
+      "step": 1600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994563080551842e-05,
+      "loss": 17.7595,
+      "step": 1601
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994542112663508e-05,
+      "loss": 17.6143,
+      "step": 1602
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99452110444291e-05,
+      "loss": 16.8653,
+      "step": 1603
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994500055890221e-05,
+      "loss": 16.0671,
+      "step": 1604
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99447896700561e-05,
+      "loss": 18.7485,
+      "step": 1605
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994457837789249e-05,
+      "loss": 15.5156,
+      "step": 1606
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994436668241303e-05,
+      "loss": 18.0699,
+      "step": 1607
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99441545836195e-05,
+      "loss": 18.7224,
+      "step": 1608
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994394208151356e-05,
+      "loss": 18.1275,
+      "step": 1609
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994372917609696e-05,
+      "loss": 17.8477,
+      "step": 1610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99435158673714e-05,
+      "loss": 17.288,
+      "step": 1611
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994330215533862e-05,
+      "loss": 16.4952,
+      "step": 1612
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994308804000033e-05,
+      "loss": 18.5924,
+      "step": 1613
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994287352135825e-05,
+      "loss": 16.9101,
+      "step": 1614
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994265859941413e-05,
+      "loss": 18.283,
+      "step": 1615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99424432741697e-05,
+      "loss": 16.5227,
+      "step": 1616
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.99422275456267e-05,
+      "loss": 17.2415,
+      "step": 1617
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994201141378688e-05,
+      "loss": 17.424,
+      "step": 1618
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994179487865198e-05,
+      "loss": 16.3578,
+      "step": 1619
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994157794022374e-05,
+      "loss": 16.9543,
+      "step": 1620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.994136059850392e-05,
+      "loss": 18.1674,
+      "step": 1621
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.994114285349426e-05,
+      "loss": 20.5325,
+      "step": 1622
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.994092470519654e-05,
+      "loss": 21.1457,
+      "step": 1623
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.994070615361252e-05,
+      "loss": 18.8358,
+      "step": 1624
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.994048719874395e-05,
+      "loss": 15.0926,
+      "step": 1625
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.994026784059261e-05,
+      "loss": 16.3136,
+      "step": 1626
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.994004807916026e-05,
+      "loss": 20.4934,
+      "step": 1627
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993982791444869e-05,
+      "loss": 19.5012,
+      "step": 1628
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993960734645966e-05,
+      "loss": 16.517,
+      "step": 1629
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993938637519497e-05,
+      "loss": 18.2267,
+      "step": 1630
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993916500065639e-05,
+      "loss": 18.6503,
+      "step": 1631
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993894322284571e-05,
+      "loss": 18.0504,
+      "step": 1632
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993872104176474e-05,
+      "loss": 16.0982,
+      "step": 1633
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993849845741524e-05,
+      "loss": 16.9338,
+      "step": 1634
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993827546979903e-05,
+      "loss": 16.0276,
+      "step": 1635
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993805207891791e-05,
+      "loss": 16.1952,
+      "step": 1636
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993782828477368e-05,
+      "loss": 16.2459,
+      "step": 1637
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993760408736814e-05,
+      "loss": 15.7505,
+      "step": 1638
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993737948670312e-05,
+      "loss": 16.7287,
+      "step": 1639
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993715448278041e-05,
+      "loss": 17.5778,
+      "step": 1640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993692907560185e-05,
+      "loss": 18.4794,
+      "step": 1641
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993670326516924e-05,
+      "loss": 17.7057,
+      "step": 1642
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993647705148442e-05,
+      "loss": 15.1103,
+      "step": 1643
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99362504345492e-05,
+      "loss": 20.1185,
+      "step": 1644
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99360234143654e-05,
+      "loss": 15.1423,
+      "step": 1645
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99357959909349e-05,
+      "loss": 19.7281,
+      "step": 1646
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993556816425951e-05,
+      "loss": 19.6157,
+      "step": 1647
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993533993434105e-05,
+      "loss": 15.9601,
+      "step": 1648
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99351113011814e-05,
+      "loss": 16.8498,
+      "step": 1649
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993488226478239e-05,
+      "loss": 18.3598,
+      "step": 1650
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993465282514585e-05,
+      "loss": 17.3618,
+      "step": 1651
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993442298227365e-05,
+      "loss": 18.4678,
+      "step": 1652
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993419273616765e-05,
+      "loss": 17.7644,
+      "step": 1653
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99339620868297e-05,
+      "loss": 18.1606,
+      "step": 1654
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99337310342617e-05,
+      "loss": 16.6552,
+      "step": 1655
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993349957846545e-05,
+      "loss": 17.4562,
+      "step": 1656
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993326771944284e-05,
+      "loss": 17.773,
+      "step": 1657
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99330354571958e-05,
+      "loss": 18.2802,
+      "step": 1658
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993280279172613e-05,
+      "loss": 17.7254,
+      "step": 1659
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993256972303573e-05,
+      "loss": 17.3497,
+      "step": 1660
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993233625112653e-05,
+      "loss": 18.1434,
+      "step": 1661
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993210237600034e-05,
+      "loss": 18.4539,
+      "step": 1662
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993186809765911e-05,
+      "loss": 18.8266,
+      "step": 1663
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99316334161047e-05,
+      "loss": 18.8687,
+      "step": 1664
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.9931398331339e-05,
+      "loss": 16.9192,
+      "step": 1665
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993116284336392e-05,
+      "loss": 18.7781,
+      "step": 1666
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993092695218138e-05,
+      "loss": 18.1294,
+      "step": 1667
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993069065779326e-05,
+      "loss": 16.8809,
+      "step": 1668
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993045396020147e-05,
+      "loss": 19.5389,
+      "step": 1669
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.993021685940793e-05,
+      "loss": 18.6702,
+      "step": 1670
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992997935541457e-05,
+      "loss": 18.5631,
+      "step": 1671
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992974144822326e-05,
+      "loss": 16.8921,
+      "step": 1672
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992950313783597e-05,
+      "loss": 16.2825,
+      "step": 1673
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99292644242546e-05,
+      "loss": 17.941,
+      "step": 1674
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992902530748107e-05,
+      "loss": 15.9985,
+      "step": 1675
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992878578751734e-05,
+      "loss": 17.423,
+      "step": 1676
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992854586436533e-05,
+      "loss": 16.672,
+      "step": 1677
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992830553802696e-05,
+      "loss": 17.7638,
+      "step": 1678
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992806480850421e-05,
+      "loss": 19.3624,
+      "step": 1679
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992782367579899e-05,
+      "loss": 17.4022,
+      "step": 1680
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992758213991324e-05,
+      "loss": 17.1257,
+      "step": 1681
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992734020084893e-05,
+      "loss": 17.7719,
+      "step": 1682
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992709785860802e-05,
+      "loss": 18.2994,
+      "step": 1683
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992685511319246e-05,
+      "loss": 18.0988,
+      "step": 1684
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99266119646042e-05,
+      "loss": 16.6398,
+      "step": 1685
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992636841284523e-05,
+      "loss": 18.5357,
+      "step": 1686
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992612445791749e-05,
+      "loss": 18.5573,
+      "step": 1687
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992588009982295e-05,
+      "loss": 16.0142,
+      "step": 1688
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99256353385636e-05,
+      "loss": 19.0321,
+      "step": 1689
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99253901741414e-05,
+      "loss": 16.7765,
+      "step": 1690
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992514460655835e-05,
+      "loss": 18.3545,
+      "step": 1691
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992489863581642e-05,
+      "loss": 18.223,
+      "step": 1692
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992465226191759e-05,
+      "loss": 17.4422,
+      "step": 1693
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992440548486385e-05,
+      "loss": 19.0459,
+      "step": 1694
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992415830465723e-05,
+      "loss": 17.634,
+      "step": 1695
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992391072129967e-05,
+      "loss": 17.2958,
+      "step": 1696
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99236627347932e-05,
+      "loss": 15.9615,
+      "step": 1697
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992341434513982e-05,
+      "loss": 16.0477,
+      "step": 1698
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992316555234154e-05,
+      "loss": 16.4711,
+      "step": 1699
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992291635640034e-05,
+      "loss": 18.1502,
+      "step": 1700
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992266675731827e-05,
+      "loss": 17.5688,
+      "step": 1701
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992241675509735e-05,
+      "loss": 17.6676,
+      "step": 1702
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992216634973955e-05,
+      "loss": 17.3971,
+      "step": 1703
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992191554124694e-05,
+      "loss": 18.4685,
+      "step": 1704
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992166432962151e-05,
+      "loss": 17.0605,
+      "step": 1705
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992141271486533e-05,
+      "loss": 19.6637,
+      "step": 1706
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992116069698039e-05,
+      "loss": 16.2974,
+      "step": 1707
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992090827596874e-05,
+      "loss": 17.2508,
+      "step": 1708
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992065545183242e-05,
+      "loss": 18.5777,
+      "step": 1709
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992040222457348e-05,
+      "loss": 17.3009,
+      "step": 1710
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.992014859419395e-05,
+      "loss": 17.5439,
+      "step": 1711
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99198945606959e-05,
+      "loss": 16.6413,
+      "step": 1712
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991964012408135e-05,
+      "loss": 16.1329,
+      "step": 1713
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991938528435239e-05,
+      "loss": 17.1491,
+      "step": 1714
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991913004151104e-05,
+      "loss": 16.9845,
+      "step": 1715
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991887439555938e-05,
+      "loss": 19.3505,
+      "step": 1716
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991861834649949e-05,
+      "loss": 19.5797,
+      "step": 1717
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991836189433342e-05,
+      "loss": 18.3408,
+      "step": 1718
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991810503906324e-05,
+      "loss": 15.9379,
+      "step": 1719
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991784778069103e-05,
+      "loss": 18.5991,
+      "step": 1720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991759011921886e-05,
+      "loss": 16.9526,
+      "step": 1721
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991733205464882e-05,
+      "loss": 17.6945,
+      "step": 1722
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.9917073586983e-05,
+      "loss": 16.2653,
+      "step": 1723
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991681471622345e-05,
+      "loss": 16.7374,
+      "step": 1724
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991655544237231e-05,
+      "loss": 16.7523,
+      "step": 1725
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991629576543163e-05,
+      "loss": 17.3061,
+      "step": 1726
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991603568540355e-05,
+      "loss": 18.8926,
+      "step": 1727
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991577520229014e-05,
+      "loss": 18.2192,
+      "step": 1728
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99155143160935e-05,
+      "loss": 17.5008,
+      "step": 1729
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991525302681578e-05,
+      "loss": 15.5953,
+      "step": 1730
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991499133445903e-05,
+      "loss": 15.4608,
+      "step": 1731
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.99147292390254e-05,
+      "loss": 17.7911,
+      "step": 1732
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991446674051698e-05,
+      "loss": 20.0426,
+      "step": 1733
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991420383893594e-05,
+      "loss": 17.7746,
+      "step": 1734
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991394053428436e-05,
+      "loss": 17.7412,
+      "step": 1735
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991367682656438e-05,
+      "loss": 16.4778,
+      "step": 1736
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991341271577812e-05,
+      "loss": 18.5211,
+      "step": 1737
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991314820192772e-05,
+      "loss": 16.874,
+      "step": 1738
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991288328501531e-05,
+      "loss": 15.7783,
+      "step": 1739
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991261796504304e-05,
+      "loss": 16.9693,
+      "step": 1740
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991235224201305e-05,
+      "loss": 18.0912,
+      "step": 1741
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.991208611592749e-05,
+      "loss": 17.8267,
+      "step": 1742
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.991181958678848e-05,
+      "loss": 17.6237,
+      "step": 1743
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.991155265459822e-05,
+      "loss": 17.0616,
+      "step": 1744
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.991128531935882e-05,
+      "loss": 17.3129,
+      "step": 1745
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.991101758107246e-05,
+      "loss": 19.4338,
+      "step": 1746
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.991074943974131e-05,
+      "loss": 17.946,
+      "step": 1747
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.991048089536752e-05,
+      "loss": 17.4321,
+      "step": 1748
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.991021194795327e-05,
+      "loss": 18.0588,
+      "step": 1749
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990994259750073e-05,
+      "loss": 17.5411,
+      "step": 1750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990967284401207e-05,
+      "loss": 17.8762,
+      "step": 1751
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990940268748947e-05,
+      "loss": 20.4666,
+      "step": 1752
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990913212793511e-05,
+      "loss": 17.0464,
+      "step": 1753
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990886116535117e-05,
+      "loss": 18.8624,
+      "step": 1754
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990858979973984e-05,
+      "loss": 19.5404,
+      "step": 1755
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990831803110332e-05,
+      "loss": 17.588,
+      "step": 1756
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990804585944381e-05,
+      "loss": 19.4009,
+      "step": 1757
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990777328476348e-05,
+      "loss": 16.3749,
+      "step": 1758
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990750030706457e-05,
+      "loss": 15.8359,
+      "step": 1759
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990722692634926e-05,
+      "loss": 18.4399,
+      "step": 1760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990695314261975e-05,
+      "loss": 16.9848,
+      "step": 1761
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990667895587827e-05,
+      "loss": 17.9111,
+      "step": 1762
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990640436612704e-05,
+      "loss": 19.9023,
+      "step": 1763
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990612937336824e-05,
+      "loss": 15.8794,
+      "step": 1764
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990585397760413e-05,
+      "loss": 17.3299,
+      "step": 1765
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.99055781788369e-05,
+      "loss": 18.3838,
+      "step": 1766
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990530197706883e-05,
+      "loss": 15.6408,
+      "step": 1767
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990502537230208e-05,
+      "loss": 16.7407,
+      "step": 1768
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990474836453894e-05,
+      "loss": 18.5692,
+      "step": 1769
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990447095378161e-05,
+      "loss": 18.6963,
+      "step": 1770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990419314003235e-05,
+      "loss": 18.0146,
+      "step": 1771
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990391492329341e-05,
+      "loss": 17.4196,
+      "step": 1772
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990363630356703e-05,
+      "loss": 16.9323,
+      "step": 1773
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990335728085545e-05,
+      "loss": 20.3598,
+      "step": 1774
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990307785516093e-05,
+      "loss": 16.8194,
+      "step": 1775
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990279802648572e-05,
+      "loss": 16.8949,
+      "step": 1776
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990251779483208e-05,
+      "loss": 18.702,
+      "step": 1777
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990223716020229e-05,
+      "loss": 17.2678,
+      "step": 1778
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.99019561225986e-05,
+      "loss": 16.6301,
+      "step": 1779
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.99016746820233e-05,
+      "loss": 17.5433,
+      "step": 1780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990139283847863e-05,
+      "loss": 17.7147,
+      "step": 1781
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990111059196689e-05,
+      "loss": 18.0988,
+      "step": 1782
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990082794249037e-05,
+      "loss": 17.1115,
+      "step": 1783
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990054489005131e-05,
+      "loss": 15.9139,
+      "step": 1784
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.990026143465203e-05,
+      "loss": 16.0003,
+      "step": 1785
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.98999775762948e-05,
+      "loss": 18.014,
+      "step": 1786
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989969331498192e-05,
+      "loss": 19.5474,
+      "step": 1787
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.98994086507157e-05,
+      "loss": 18.006,
+      "step": 1788
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989912358349842e-05,
+      "loss": 19.7903,
+      "step": 1789
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989883811333241e-05,
+      "loss": 17.7332,
+      "step": 1790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989855224021993e-05,
+      "loss": 19.5094,
+      "step": 1791
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.98982659641633e-05,
+      "loss": 16.5504,
+      "step": 1792
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989797928516486e-05,
+      "loss": 19.5113,
+      "step": 1793
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989769220322692e-05,
+      "loss": 18.3515,
+      "step": 1794
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989740471835176e-05,
+      "loss": 17.972,
+      "step": 1795
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989711683054175e-05,
+      "loss": 18.2796,
+      "step": 1796
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989682853979918e-05,
+      "loss": 15.6113,
+      "step": 1797
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.98965398461264e-05,
+      "loss": 17.0204,
+      "step": 1798
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989625074952574e-05,
+      "loss": 17.1472,
+      "step": 1799
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.98959612499995e-05,
+      "loss": 18.1255,
+      "step": 1800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989567134755007e-05,
+      "loss": 16.8282,
+      "step": 1801
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989538104217975e-05,
+      "loss": 16.0084,
+      "step": 1802
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989509033389091e-05,
+      "loss": 18.4625,
+      "step": 1803
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989479922268588e-05,
+      "loss": 17.0461,
+      "step": 1804
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989450770856703e-05,
+      "loss": 17.3487,
+      "step": 1805
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989421579153669e-05,
+      "loss": 18.2362,
+      "step": 1806
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989392347159723e-05,
+      "loss": 17.3199,
+      "step": 1807
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989363074875103e-05,
+      "loss": 16.7758,
+      "step": 1808
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989333762300041e-05,
+      "loss": 16.928,
+      "step": 1809
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989304409434776e-05,
+      "loss": 17.3441,
+      "step": 1810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989275016279547e-05,
+      "loss": 16.5877,
+      "step": 1811
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989245582834589e-05,
+      "loss": 18.4359,
+      "step": 1812
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989216109100139e-05,
+      "loss": 18.0721,
+      "step": 1813
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989186595076438e-05,
+      "loss": 15.9421,
+      "step": 1814
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989157040763722e-05,
+      "loss": 15.5934,
+      "step": 1815
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989127446162229e-05,
+      "loss": 18.9047,
+      "step": 1816
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.9890978112722e-05,
+      "loss": 17.5532,
+      "step": 1817
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989068136093873e-05,
+      "loss": 17.4926,
+      "step": 1818
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989038420627489e-05,
+      "loss": 16.6216,
+      "step": 1819
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.989008664873287e-05,
+      "loss": 18.6748,
+      "step": 1820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988978868831506e-05,
+      "loss": 17.1954,
+      "step": 1821
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988949032502389e-05,
+      "loss": 18.7437,
+      "step": 1822
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988919155886176e-05,
+      "loss": 16.4463,
+      "step": 1823
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988889238983107e-05,
+      "loss": 18.8644,
+      "step": 1824
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988859281793428e-05,
+      "loss": 18.6354,
+      "step": 1825
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988829284317375e-05,
+      "loss": 17.0895,
+      "step": 1826
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988799246555193e-05,
+      "loss": 19.0776,
+      "step": 1827
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988769168507125e-05,
+      "loss": 17.1903,
+      "step": 1828
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988739050173414e-05,
+      "loss": 15.6281,
+      "step": 1829
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988708891554302e-05,
+      "loss": 18.897,
+      "step": 1830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988678692650033e-05,
+      "loss": 17.0725,
+      "step": 1831
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.98864845346085e-05,
+      "loss": 18.3419,
+      "step": 1832
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988618173986999e-05,
+      "loss": 16.3487,
+      "step": 1833
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988587854228724e-05,
+      "loss": 18.0836,
+      "step": 1834
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.98855749418627e-05,
+      "loss": 17.3431,
+      "step": 1835
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988527093859881e-05,
+      "loss": 17.3369,
+      "step": 1836
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988496653249802e-05,
+      "loss": 17.5436,
+      "step": 1837
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988466172356282e-05,
+      "loss": 16.5729,
+      "step": 1838
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988435651179564e-05,
+      "loss": 15.6238,
+      "step": 1839
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988405089719897e-05,
+      "loss": 16.2125,
+      "step": 1840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988374487977526e-05,
+      "loss": 18.859,
+      "step": 1841
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988343845952697e-05,
+      "loss": 18.0713,
+      "step": 1842
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.98831316364566e-05,
+      "loss": 16.7451,
+      "step": 1843
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.98828244105666e-05,
+      "loss": 19.3246,
+      "step": 1844
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.98825167818595e-05,
+      "loss": 15.8054,
+      "step": 1845
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988220875033774e-05,
+      "loss": 17.7313,
+      "step": 1846
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.98819003160038e-05,
+      "loss": 16.2044,
+      "step": 1847
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988159147886021e-05,
+      "loss": 19.8754,
+      "step": 1848
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988128223890945e-05,
+      "loss": 17.0064,
+      "step": 1849
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988097259615399e-05,
+      "loss": 17.4306,
+      "step": 1850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988066255059636e-05,
+      "loss": 16.7988,
+      "step": 1851
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.988035210223907e-05,
+      "loss": 16.4375,
+      "step": 1852
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.98800412510846e-05,
+      "loss": 18.568,
+      "step": 1853
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.987972999713549e-05,
+      "loss": 17.4415,
+      "step": 1854
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.987941834039421e-05,
+      "loss": 17.4762,
+      "step": 1855
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.987910628086333e-05,
+      "loss": 17.0352,
+      "step": 1856
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.987879381854532e-05,
+      "loss": 16.852,
+      "step": 1857
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.987848095344272e-05,
+      "loss": 17.5047,
+      "step": 1858
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.987816768555809e-05,
+      "loss": 16.7678,
+      "step": 1859
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.987785401489391e-05,
+      "loss": 19.0492,
+      "step": 1860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.987753994145275e-05,
+      "loss": 16.7629,
+      "step": 1861
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.987722546523712e-05,
+      "loss": 21.538,
+      "step": 1862
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98769105862496e-05,
+      "loss": 17.4908,
+      "step": 1863
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987659530449268e-05,
+      "loss": 17.0072,
+      "step": 1864
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987627961996892e-05,
+      "loss": 16.7743,
+      "step": 1865
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987596353268089e-05,
+      "loss": 16.9345,
+      "step": 1866
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987564704263113e-05,
+      "loss": 16.3059,
+      "step": 1867
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98753301498222e-05,
+      "loss": 17.6184,
+      "step": 1868
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987501285425667e-05,
+      "loss": 18.6798,
+      "step": 1869
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987469515593707e-05,
+      "loss": 17.2616,
+      "step": 1870
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987437705486598e-05,
+      "loss": 18.0287,
+      "step": 1871
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.9874058551046e-05,
+      "loss": 15.8429,
+      "step": 1872
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987373964447965e-05,
+      "loss": 15.2432,
+      "step": 1873
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987342033516953e-05,
+      "loss": 18.2741,
+      "step": 1874
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987310062311823e-05,
+      "loss": 16.8603,
+      "step": 1875
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98727805083283e-05,
+      "loss": 18.9514,
+      "step": 1876
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987245999080237e-05,
+      "loss": 15.4372,
+      "step": 1877
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.9872139070543e-05,
+      "loss": 16.7489,
+      "step": 1878
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987181774755278e-05,
+      "loss": 15.8248,
+      "step": 1879
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98714960218343e-05,
+      "loss": 18.5889,
+      "step": 1880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987117389339019e-05,
+      "loss": 16.4889,
+      "step": 1881
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987085136222302e-05,
+      "loss": 17.1999,
+      "step": 1882
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98705284283354e-05,
+      "loss": 16.9199,
+      "step": 1883
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987020509172994e-05,
+      "loss": 16.9916,
+      "step": 1884
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986988135240926e-05,
+      "loss": 17.2851,
+      "step": 1885
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986955721037596e-05,
+      "loss": 15.6192,
+      "step": 1886
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986923266563267e-05,
+      "loss": 16.1025,
+      "step": 1887
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.9868907718182e-05,
+      "loss": 17.8784,
+      "step": 1888
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986858236802658e-05,
+      "loss": 18.0052,
+      "step": 1889
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986825661516903e-05,
+      "loss": 17.8769,
+      "step": 1890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.9867930459612e-05,
+      "loss": 16.1547,
+      "step": 1891
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98676039013581e-05,
+      "loss": 16.3938,
+      "step": 1892
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986727694040999e-05,
+      "loss": 17.8983,
+      "step": 1893
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986694957677029e-05,
+      "loss": 18.6054,
+      "step": 1894
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986662181044165e-05,
+      "loss": 18.8438,
+      "step": 1895
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986629364142672e-05,
+      "loss": 17.1612,
+      "step": 1896
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986596506972815e-05,
+      "loss": 17.255,
+      "step": 1897
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986563609534859e-05,
+      "loss": 16.6476,
+      "step": 1898
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98653067182907e-05,
+      "loss": 17.0537,
+      "step": 1899
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986497693855714e-05,
+      "loss": 17.1977,
+      "step": 1900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986464675615057e-05,
+      "loss": 18.5329,
+      "step": 1901
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986431617107366e-05,
+      "loss": 17.598,
+      "step": 1902
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986398518332907e-05,
+      "loss": 19.7377,
+      "step": 1903
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986365379291949e-05,
+      "loss": 17.0438,
+      "step": 1904
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986332199984758e-05,
+      "loss": 17.2001,
+      "step": 1905
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986298980411603e-05,
+      "loss": 19.194,
+      "step": 1906
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98626572057275e-05,
+      "loss": 17.815,
+      "step": 1907
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98623242046847e-05,
+      "loss": 17.731,
+      "step": 1908
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986199080099034e-05,
+      "loss": 17.6279,
+      "step": 1909
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986165699464705e-05,
+      "loss": 17.8272,
+      "step": 1910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986132278565759e-05,
+      "loss": 17.5373,
+      "step": 1911
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98609881740246e-05,
+      "loss": 16.4515,
+      "step": 1912
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986065315975084e-05,
+      "loss": 15.8001,
+      "step": 1913
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.986031774283897e-05,
+      "loss": 17.2905,
+      "step": 1914
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985998192329172e-05,
+      "loss": 18.7255,
+      "step": 1915
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98596457011118e-05,
+      "loss": 18.1574,
+      "step": 1916
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985930907630193e-05,
+      "loss": 18.6401,
+      "step": 1917
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985897204886481e-05,
+      "loss": 17.3444,
+      "step": 1918
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985863461880316e-05,
+      "loss": 17.8901,
+      "step": 1919
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985829678611973e-05,
+      "loss": 16.9821,
+      "step": 1920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985795855081723e-05,
+      "loss": 17.9419,
+      "step": 1921
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985761991289842e-05,
+      "loss": 18.2916,
+      "step": 1922
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985728087236599e-05,
+      "loss": 16.2271,
+      "step": 1923
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985694142922269e-05,
+      "loss": 17.082,
+      "step": 1924
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985660158347128e-05,
+      "loss": 18.5299,
+      "step": 1925
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985626133511449e-05,
+      "loss": 18.766,
+      "step": 1926
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98559206841551e-05,
+      "loss": 16.6524,
+      "step": 1927
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985557963059581e-05,
+      "loss": 18.6962,
+      "step": 1928
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98552381744394e-05,
+      "loss": 19.2405,
+      "step": 1929
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985489631568863e-05,
+      "loss": 17.339,
+      "step": 1930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985455405434624e-05,
+      "loss": 17.9968,
+      "step": 1931
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985421139041502e-05,
+      "loss": 17.8613,
+      "step": 1932
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985386832389771e-05,
+      "loss": 20.0336,
+      "step": 1933
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985352485479712e-05,
+      "loss": 17.6123,
+      "step": 1934
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.9853180983116e-05,
+      "loss": 16.7839,
+      "step": 1935
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98528367088571e-05,
+      "loss": 19.0589,
+      "step": 1936
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985249203202324e-05,
+      "loss": 15.3337,
+      "step": 1937
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98521469526172e-05,
+      "loss": 17.379,
+      "step": 1938
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985180147064174e-05,
+      "loss": 18.2169,
+      "step": 1939
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985145558609969e-05,
+      "loss": 16.2804,
+      "step": 1940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985110929899381e-05,
+      "loss": 16.485,
+      "step": 1941
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985076260932691e-05,
+      "loss": 14.9935,
+      "step": 1942
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985041551710177e-05,
+      "loss": 18.0153,
+      "step": 1943
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.985006802232122e-05,
+      "loss": 18.5125,
+      "step": 1944
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984972012498805e-05,
+      "loss": 16.2699,
+      "step": 1945
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984937182510507e-05,
+      "loss": 19.6704,
+      "step": 1946
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98490231226751e-05,
+      "loss": 16.6008,
+      "step": 1947
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984867401770097e-05,
+      "loss": 15.3239,
+      "step": 1948
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984832451018547e-05,
+      "loss": 17.5771,
+      "step": 1949
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984797460013143e-05,
+      "loss": 16.2954,
+      "step": 1950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984762428754167e-05,
+      "loss": 17.5326,
+      "step": 1951
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984727357241905e-05,
+      "loss": 18.1069,
+      "step": 1952
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984692245476637e-05,
+      "loss": 19.0241,
+      "step": 1953
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984657093458647e-05,
+      "loss": 20.712,
+      "step": 1954
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98462190118822e-05,
+      "loss": 14.8945,
+      "step": 1955
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98458666866564e-05,
+      "loss": 15.7464,
+      "step": 1956
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98455139589119e-05,
+      "loss": 18.7495,
+      "step": 1957
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984516082865159e-05,
+      "loss": 17.7293,
+      "step": 1958
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984480729587826e-05,
+      "loss": 16.6044,
+      "step": 1959
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984445336059481e-05,
+      "loss": 19.5083,
+      "step": 1960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984409902280408e-05,
+      "loss": 17.1936,
+      "step": 1961
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984374428250894e-05,
+      "loss": 20.1659,
+      "step": 1962
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984338913971225e-05,
+      "loss": 16.5485,
+      "step": 1963
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984303359441688e-05,
+      "loss": 17.8579,
+      "step": 1964
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984267764662569e-05,
+      "loss": 17.3901,
+      "step": 1965
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984232129634158e-05,
+      "loss": 15.4222,
+      "step": 1966
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984196454356741e-05,
+      "loss": 17.5981,
+      "step": 1967
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984160738830604e-05,
+      "loss": 17.6413,
+      "step": 1968
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98412498305604e-05,
+      "loss": 16.9297,
+      "step": 1969
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984089187033334e-05,
+      "loss": 17.0387,
+      "step": 1970
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984053350762778e-05,
+      "loss": 16.1684,
+      "step": 1971
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.984017474244659e-05,
+      "loss": 19.4224,
+      "step": 1972
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.983981557479267e-05,
+      "loss": 17.0818,
+      "step": 1973
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.983945600466893e-05,
+      "loss": 16.1628,
+      "step": 1974
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.983909603207829e-05,
+      "loss": 19.8037,
+      "step": 1975
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.98387356570236e-05,
+      "loss": 18.3465,
+      "step": 1976
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.983837487950782e-05,
+      "loss": 16.2481,
+      "step": 1977
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.983801369953386e-05,
+      "loss": 16.6878,
+      "step": 1978
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.983765211710463e-05,
+      "loss": 15.336,
+      "step": 1979
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.983729013222303e-05,
+      "loss": 18.056,
+      "step": 1980
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.9836927744892e-05,
+      "loss": 18.1917,
+      "step": 1981
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.983656495511448e-05,
+      "loss": 18.4097,
+      "step": 1982
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983620176289338e-05,
+      "loss": 17.4223,
+      "step": 1983
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983583816823163e-05,
+      "loss": 18.647,
+      "step": 1984
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983547417113219e-05,
+      "loss": 18.5804,
+      "step": 1985
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983510977159798e-05,
+      "loss": 17.9634,
+      "step": 1986
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983474496963194e-05,
+      "loss": 17.9226,
+      "step": 1987
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983437976523702e-05,
+      "loss": 16.4361,
+      "step": 1988
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983401415841617e-05,
+      "loss": 18.7558,
+      "step": 1989
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983364814917236e-05,
+      "loss": 16.9554,
+      "step": 1990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983328173750852e-05,
+      "loss": 17.3813,
+      "step": 1991
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983291492342762e-05,
+      "loss": 17.5995,
+      "step": 1992
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983254770693263e-05,
+      "loss": 15.4216,
+      "step": 1993
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983218008802648e-05,
+      "loss": 16.1091,
+      "step": 1994
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983181206671219e-05,
+      "loss": 18.4671,
+      "step": 1995
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983144364299269e-05,
+      "loss": 16.5397,
+      "step": 1996
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983107481687097e-05,
+      "loss": 18.3042,
+      "step": 1997
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983070558835002e-05,
+      "loss": 18.7524,
+      "step": 1998
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.983033595743281e-05,
+      "loss": 16.4627,
+      "step": 1999
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982996592412233e-05,
+      "loss": 16.0467,
+      "step": 2000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982959548842156e-05,
+      "loss": 17.9571,
+      "step": 2001
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.98292246503335e-05,
+      "loss": 17.7982,
+      "step": 2002
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982885340986114e-05,
+      "loss": 16.8982,
+      "step": 2003
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982848176700748e-05,
+      "loss": 17.9653,
+      "step": 2004
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.98281097217755e-05,
+      "loss": 16.2943,
+      "step": 2005
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982773727416825e-05,
+      "loss": 18.2717,
+      "step": 2006
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982736442418869e-05,
+      "loss": 16.323,
+      "step": 2007
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982699117183986e-05,
+      "loss": 16.999,
+      "step": 2008
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982661751712477e-05,
+      "loss": 16.1895,
+      "step": 2009
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982624346004644e-05,
+      "loss": 18.607,
+      "step": 2010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982586900060788e-05,
+      "loss": 15.6785,
+      "step": 2011
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982549413881212e-05,
+      "loss": 18.3865,
+      "step": 2012
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982511887466218e-05,
+      "loss": 18.063,
+      "step": 2013
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982474320816108e-05,
+      "loss": 18.1576,
+      "step": 2014
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.98243671393119e-05,
+      "loss": 17.6913,
+      "step": 2015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982399066811763e-05,
+      "loss": 15.0531,
+      "step": 2016
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982361379458133e-05,
+      "loss": 18.0385,
+      "step": 2017
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982323651870604e-05,
+      "loss": 15.8278,
+      "step": 2018
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982285884049481e-05,
+      "loss": 17.341,
+      "step": 2019
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982248075995068e-05,
+      "loss": 18.2548,
+      "step": 2020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982210227707671e-05,
+      "loss": 16.8658,
+      "step": 2021
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982172339187596e-05,
+      "loss": 18.3903,
+      "step": 2022
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982134410435148e-05,
+      "loss": 14.4542,
+      "step": 2023
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982096441450635e-05,
+      "loss": 19.5082,
+      "step": 2024
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.98205843223436e-05,
+      "loss": 20.1085,
+      "step": 2025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.982020382786635e-05,
+      "loss": 18.3111,
+      "step": 2026
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981982293107764e-05,
+      "loss": 15.7705,
+      "step": 2027
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981944163198054e-05,
+      "loss": 15.8228,
+      "step": 2028
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981905993057815e-05,
+      "loss": 18.7062,
+      "step": 2029
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981867782687353e-05,
+      "loss": 17.047,
+      "step": 2030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.98182953208698e-05,
+      "loss": 18.3713,
+      "step": 2031
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981791241257e-05,
+      "loss": 18.1728,
+      "step": 2032
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981752910197728e-05,
+      "loss": 17.085,
+      "step": 2033
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981714538909468e-05,
+      "loss": 15.0528,
+      "step": 2034
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981676127392534e-05,
+      "loss": 18.273,
+      "step": 2035
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981637675647234e-05,
+      "loss": 16.8614,
+      "step": 2036
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.98159918367388e-05,
+      "loss": 18.047,
+      "step": 2037
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981560651472781e-05,
+      "loss": 16.827,
+      "step": 2038
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.98152207904425e-05,
+      "loss": 18.8955,
+      "step": 2039
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981483466388594e-05,
+      "loss": 16.6344,
+      "step": 2040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981444813506133e-05,
+      "loss": 18.6818,
+      "step": 2041
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981406120397172e-05,
+      "loss": 17.3016,
+      "step": 2042
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981367387062026e-05,
+      "loss": 16.637,
+      "step": 2043
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981328613501008e-05,
+      "loss": 14.8597,
+      "step": 2044
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981289799714432e-05,
+      "loss": 18.2993,
+      "step": 2045
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.98125094570261e-05,
+      "loss": 15.7387,
+      "step": 2046
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981212051465854e-05,
+      "loss": 17.5163,
+      "step": 2047
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981173117004484e-05,
+      "loss": 16.7878,
+      "step": 2048
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981134142318806e-05,
+      "loss": 17.239,
+      "step": 2049
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981095127409141e-05,
+      "loss": 19.1966,
+      "step": 2050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981056072275803e-05,
+      "loss": 17.0236,
+      "step": 2051
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981016976919106e-05,
+      "loss": 15.885,
+      "step": 2052
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980977841339367e-05,
+      "loss": 17.5522,
+      "step": 2053
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980938665536901e-05,
+      "loss": 17.135,
+      "step": 2054
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980899449512025e-05,
+      "loss": 18.3485,
+      "step": 2055
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980860193265056e-05,
+      "loss": 17.805,
+      "step": 2056
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.98082089679631e-05,
+      "loss": 16.1505,
+      "step": 2057
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980781560106103e-05,
+      "loss": 18.3809,
+      "step": 2058
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980742183194756e-05,
+      "loss": 18.4641,
+      "step": 2059
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980702766062587e-05,
+      "loss": 16.5349,
+      "step": 2060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.98066330870991e-05,
+      "loss": 17.8323,
+      "step": 2061
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980623811137046e-05,
+      "loss": 16.4386,
+      "step": 2062
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980584273344317e-05,
+      "loss": 16.9398,
+      "step": 2063
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980544695332038e-05,
+      "loss": 15.8889,
+      "step": 2064
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.98050507710053e-05,
+      "loss": 20.0535,
+      "step": 2065
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980465418650113e-05,
+      "loss": 18.3975,
+      "step": 2066
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980425719981109e-05,
+      "loss": 16.4841,
+      "step": 2067
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980385981093836e-05,
+      "loss": 18.5935,
+      "step": 2068
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980346201988616e-05,
+      "loss": 17.2017,
+      "step": 2069
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.98030638266577e-05,
+      "loss": 17.8776,
+      "step": 2070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.98026652312562e-05,
+      "loss": 16.8888,
+      "step": 2071
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980226623368486e-05,
+      "loss": 18.4917,
+      "step": 2072
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980186683394692e-05,
+      "loss": 19.7799,
+      "step": 2073
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980146703204561e-05,
+      "loss": 17.9311,
+      "step": 2074
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980106682798415e-05,
+      "loss": 16.478,
+      "step": 2075
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980066622176577e-05,
+      "loss": 16.2216,
+      "step": 2076
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.980026521339371e-05,
+      "loss": 20.3162,
+      "step": 2077
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.97998638028712e-05,
+      "loss": 17.5878,
+      "step": 2078
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979946199020149e-05,
+      "loss": 19.0403,
+      "step": 2079
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979905977538783e-05,
+      "loss": 17.6065,
+      "step": 2080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979865715843345e-05,
+      "loss": 18.9083,
+      "step": 2081
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979825413934162e-05,
+      "loss": 16.0106,
+      "step": 2082
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979785071811559e-05,
+      "loss": 16.3627,
+      "step": 2083
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979744689475858e-05,
+      "loss": 18.0803,
+      "step": 2084
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979704266927392e-05,
+      "loss": 17.9259,
+      "step": 2085
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979663804166484e-05,
+      "loss": 16.6534,
+      "step": 2086
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979623301193458e-05,
+      "loss": 17.9395,
+      "step": 2087
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979582758008645e-05,
+      "loss": 20.3183,
+      "step": 2088
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.97954217461237e-05,
+      "loss": 16.998,
+      "step": 2089
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979501551004964e-05,
+      "loss": 16.7681,
+      "step": 2090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979460887186752e-05,
+      "loss": 17.0814,
+      "step": 2091
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979420183158064e-05,
+      "loss": 17.4972,
+      "step": 2092
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979379438919226e-05,
+      "loss": 17.5142,
+      "step": 2093
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979338654470569e-05,
+      "loss": 17.7998,
+      "step": 2094
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979297829812424e-05,
+      "loss": 17.7464,
+      "step": 2095
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979256964945118e-05,
+      "loss": 17.4273,
+      "step": 2096
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979216059868982e-05,
+      "loss": 18.6799,
+      "step": 2097
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979175114584348e-05,
+      "loss": 17.1663,
+      "step": 2098
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.97913412909154e-05,
+      "loss": 16.596,
+      "step": 2099
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979093103390899e-05,
+      "loss": 17.4979,
+      "step": 2100
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979052037482749e-05,
+      "loss": 17.3352,
+      "step": 2101
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979010931367425e-05,
+      "loss": 16.295,
+      "step": 2102
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978969785045256e-05,
+      "loss": 19.5315,
+      "step": 2103
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978928598516577e-05,
+      "loss": 16.0629,
+      "step": 2104
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978887371781719e-05,
+      "loss": 18.5889,
+      "step": 2105
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978846104841016e-05,
+      "loss": 15.5981,
+      "step": 2106
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.9788047976948e-05,
+      "loss": 16.1311,
+      "step": 2107
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978763450343407e-05,
+      "loss": 18.1234,
+      "step": 2108
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978722062787168e-05,
+      "loss": 17.8818,
+      "step": 2109
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978680635026418e-05,
+      "loss": 18.8742,
+      "step": 2110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978639167061494e-05,
+      "loss": 16.2988,
+      "step": 2111
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978597658892726e-05,
+      "loss": 16.4827,
+      "step": 2112
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978556110520454e-05,
+      "loss": 17.2436,
+      "step": 2113
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97851452194501e-05,
+      "loss": 17.7887,
+      "step": 2114
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978472893166732e-05,
+      "loss": 17.7656,
+      "step": 2115
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978431224185958e-05,
+      "loss": 18.4892,
+      "step": 2116
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97838951500302e-05,
+      "loss": 17.357,
+      "step": 2117
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978347765618257e-05,
+      "loss": 17.7162,
+      "step": 2118
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978305976032005e-05,
+      "loss": 17.4386,
+      "step": 2119
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978264146244603e-05,
+      "loss": 17.689,
+      "step": 2120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978222276256388e-05,
+      "loss": 17.1681,
+      "step": 2121
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.9781803660677e-05,
+      "loss": 17.209,
+      "step": 2122
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978138415678874e-05,
+      "loss": 17.7334,
+      "step": 2123
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978096425090251e-05,
+      "loss": 16.8504,
+      "step": 2124
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97805439430217e-05,
+      "loss": 17.6681,
+      "step": 2125
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.978012323314969e-05,
+      "loss": 16.8339,
+      "step": 2126
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97797021212899e-05,
+      "loss": 17.8417,
+      "step": 2127
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977928060744571e-05,
+      "loss": 19.2557,
+      "step": 2128
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977885869162054e-05,
+      "loss": 17.3918,
+      "step": 2129
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977843637381777e-05,
+      "loss": 19.1054,
+      "step": 2130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977801365404085e-05,
+      "loss": 16.64,
+      "step": 2131
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977759053229317e-05,
+      "loss": 17.1036,
+      "step": 2132
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977716700857814e-05,
+      "loss": 16.1663,
+      "step": 2133
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97767430828992e-05,
+      "loss": 17.5195,
+      "step": 2134
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977631875525975e-05,
+      "loss": 19.2619,
+      "step": 2135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977589402566325e-05,
+      "loss": 17.3366,
+      "step": 2136
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97754688941131e-05,
+      "loss": 15.7043,
+      "step": 2137
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977504336061275e-05,
+      "loss": 17.1926,
+      "step": 2138
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977461742516561e-05,
+      "loss": 16.5181,
+      "step": 2139
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977419108777514e-05,
+      "loss": 18.7843,
+      "step": 2140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97737643484448e-05,
+      "loss": 18.305,
+      "step": 2141
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977333720717801e-05,
+      "loss": 14.5094,
+      "step": 2142
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977290966397822e-05,
+      "loss": 17.6078,
+      "step": 2143
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97724817188489e-05,
+      "loss": 17.9702,
+      "step": 2144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977205337179348e-05,
+      "loss": 17.5511,
+      "step": 2145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977162462281544e-05,
+      "loss": 19.0986,
+      "step": 2146
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977119547191824e-05,
+      "loss": 18.8395,
+      "step": 2147
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977076591910535e-05,
+      "loss": 16.5217,
+      "step": 2148
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.977033596438022e-05,
+      "loss": 19.8008,
+      "step": 2149
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976990560774633e-05,
+      "loss": 17.1259,
+      "step": 2150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976947484920716e-05,
+      "loss": 17.2404,
+      "step": 2151
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976904368876619e-05,
+      "loss": 18.964,
+      "step": 2152
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97686121264269e-05,
+      "loss": 15.7753,
+      "step": 2153
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976818016219277e-05,
+      "loss": 17.1366,
+      "step": 2154
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976774779606728e-05,
+      "loss": 16.2344,
+      "step": 2155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976731502805395e-05,
+      "loss": 18.4997,
+      "step": 2156
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976688185815623e-05,
+      "loss": 17.393,
+      "step": 2157
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976644828637767e-05,
+      "loss": 17.5572,
+      "step": 2158
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976601431272175e-05,
+      "loss": 16.8747,
+      "step": 2159
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976557993719196e-05,
+      "loss": 15.7158,
+      "step": 2160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976514515979181e-05,
+      "loss": 20.3542,
+      "step": 2161
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976470998052483e-05,
+      "loss": 17.9155,
+      "step": 2162
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976427439939453e-05,
+      "loss": 18.1664,
+      "step": 2163
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97638384164044e-05,
+      "loss": 18.2593,
+      "step": 2164
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.9763402031558e-05,
+      "loss": 17.1092,
+      "step": 2165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976296524485884e-05,
+      "loss": 17.2322,
+      "step": 2166
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976252805631043e-05,
+      "loss": 17.2667,
+      "step": 2167
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976209046591632e-05,
+      "loss": 18.4428,
+      "step": 2168
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976165247368003e-05,
+      "loss": 16.9255,
+      "step": 2169
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976121407960511e-05,
+      "loss": 18.2069,
+      "step": 2170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976077528369509e-05,
+      "loss": 16.8794,
+      "step": 2171
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.976033608595354e-05,
+      "loss": 14.9922,
+      "step": 2172
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975989648638395e-05,
+      "loss": 15.1774,
+      "step": 2173
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975945648498992e-05,
+      "loss": 17.1914,
+      "step": 2174
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.9759016081775e-05,
+      "loss": 18.4444,
+      "step": 2175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975857527674271e-05,
+      "loss": 19.0308,
+      "step": 2176
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975813406989665e-05,
+      "loss": 17.2402,
+      "step": 2177
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975769246124035e-05,
+      "loss": 16.3145,
+      "step": 2178
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97572504507774e-05,
+      "loss": 16.0401,
+      "step": 2179
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975680803851137e-05,
+      "loss": 19.1559,
+      "step": 2180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975636522444582e-05,
+      "loss": 18.9672,
+      "step": 2181
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975592200858432e-05,
+      "loss": 15.9776,
+      "step": 2182
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975547839093046e-05,
+      "loss": 17.5197,
+      "step": 2183
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975503437148783e-05,
+      "loss": 19.2398,
+      "step": 2184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975458995026001e-05,
+      "loss": 17.4896,
+      "step": 2185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975414512725057e-05,
+      "loss": 17.4658,
+      "step": 2186
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975369990246313e-05,
+      "loss": 19.7856,
+      "step": 2187
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975325427590127e-05,
+      "loss": 15.3013,
+      "step": 2188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97528082475686e-05,
+      "loss": 16.4204,
+      "step": 2189
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97523618174687e-05,
+      "loss": 18.094,
+      "step": 2190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975191498560519e-05,
+      "loss": 16.7155,
+      "step": 2191
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975146775198168e-05,
+      "loss": 16.9119,
+      "step": 2192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975102011660179e-05,
+      "loss": 18.7998,
+      "step": 2193
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975057207946911e-05,
+      "loss": 17.2843,
+      "step": 2194
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.975012364058727e-05,
+      "loss": 17.3493,
+      "step": 2195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97496747999599e-05,
+      "loss": 19.5954,
+      "step": 2196
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974922555759061e-05,
+      "loss": 15.4766,
+      "step": 2197
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974877591348304e-05,
+      "loss": 17.9854,
+      "step": 2198
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974832586764081e-05,
+      "loss": 17.5789,
+      "step": 2199
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974787542006756e-05,
+      "loss": 19.9132,
+      "step": 2200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974742457076694e-05,
+      "loss": 17.6913,
+      "step": 2201
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974697331974256e-05,
+      "loss": 19.0089,
+      "step": 2202
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974652166699808e-05,
+      "loss": 18.7301,
+      "step": 2203
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974606961253716e-05,
+      "loss": 15.5086,
+      "step": 2204
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974561715636344e-05,
+      "loss": 16.4568,
+      "step": 2205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974516429848057e-05,
+      "loss": 17.5201,
+      "step": 2206
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974471103889222e-05,
+      "loss": 18.8211,
+      "step": 2207
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974425737760202e-05,
+      "loss": 16.333,
+      "step": 2208
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974380331461368e-05,
+      "loss": 17.0461,
+      "step": 2209
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974334884993082e-05,
+      "loss": 18.216,
+      "step": 2210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974289398355714e-05,
+      "loss": 17.9922,
+      "step": 2211
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974243871549631e-05,
+      "loss": 16.2159,
+      "step": 2212
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974198304575198e-05,
+      "loss": 17.3998,
+      "step": 2213
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974152697432787e-05,
+      "loss": 16.5788,
+      "step": 2214
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974107050122762e-05,
+      "loss": 15.5908,
+      "step": 2215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974061362645498e-05,
+      "loss": 17.255,
+      "step": 2216
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.974015635001356e-05,
+      "loss": 17.9185,
+      "step": 2217
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97396986719071e-05,
+      "loss": 17.0605,
+      "step": 2218
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.973924059213928e-05,
+      "loss": 16.9596,
+      "step": 2219
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.973878211071383e-05,
+      "loss": 15.8369,
+      "step": 2220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.973832322763441e-05,
+      "loss": 17.9516,
+      "step": 2221
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.973786394290474e-05,
+      "loss": 17.0525,
+      "step": 2222
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973740425652855e-05,
+      "loss": 19.0861,
+      "step": 2223
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973694416850951e-05,
+      "loss": 18.6045,
+      "step": 2224
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973648367885137e-05,
+      "loss": 17.9711,
+      "step": 2225
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973602278755786e-05,
+      "loss": 17.1995,
+      "step": 2226
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973556149463266e-05,
+      "loss": 15.7061,
+      "step": 2227
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973509980007953e-05,
+      "loss": 17.9563,
+      "step": 2228
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973463770390217e-05,
+      "loss": 18.5826,
+      "step": 2229
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973417520610434e-05,
+      "loss": 18.1278,
+      "step": 2230
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973371230668974e-05,
+      "loss": 16.6817,
+      "step": 2231
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973324900566213e-05,
+      "loss": 17.1638,
+      "step": 2232
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973278530302526e-05,
+      "loss": 17.0554,
+      "step": 2233
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973232119878286e-05,
+      "loss": 20.0711,
+      "step": 2234
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973185669293869e-05,
+      "loss": 15.6911,
+      "step": 2235
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973139178549649e-05,
+      "loss": 17.7697,
+      "step": 2236
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973092647646001e-05,
+      "loss": 15.6157,
+      "step": 2237
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.973046076583301e-05,
+      "loss": 17.2172,
+      "step": 2238
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972999465361927e-05,
+      "loss": 17.4277,
+      "step": 2239
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972952813982253e-05,
+      "loss": 17.6836,
+      "step": 2240
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972906122444657e-05,
+      "loss": 17.4692,
+      "step": 2241
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972859390749516e-05,
+      "loss": 15.892,
+      "step": 2242
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972812618897205e-05,
+      "loss": 16.4595,
+      "step": 2243
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972765806888106e-05,
+      "loss": 17.7707,
+      "step": 2244
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972718954722594e-05,
+      "loss": 14.6145,
+      "step": 2245
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972672062401048e-05,
+      "loss": 17.1944,
+      "step": 2246
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972625129923847e-05,
+      "loss": 16.7583,
+      "step": 2247
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.97257815729137e-05,
+      "loss": 15.8428,
+      "step": 2248
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972531144503996e-05,
+      "loss": 18.9052,
+      "step": 2249
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972484091562105e-05,
+      "loss": 18.103,
+      "step": 2250
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972436998466077e-05,
+      "loss": 17.2986,
+      "step": 2251
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972389865216292e-05,
+      "loss": 15.2522,
+      "step": 2252
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972342691813129e-05,
+      "loss": 18.0561,
+      "step": 2253
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972295478256972e-05,
+      "loss": 18.9293,
+      "step": 2254
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.9722482245482e-05,
+      "loss": 15.8753,
+      "step": 2255
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972200930687197e-05,
+      "loss": 18.9095,
+      "step": 2256
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.97215359667434e-05,
+      "loss": 17.074,
+      "step": 2257
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972106222510018e-05,
+      "loss": 16.6409,
+      "step": 2258
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972058808194608e-05,
+      "loss": 16.3617,
+      "step": 2259
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.972011353728496e-05,
+      "loss": 17.3038,
+      "step": 2260
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971963859112064e-05,
+      "loss": 15.1277,
+      "step": 2261
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971916324345696e-05,
+      "loss": 16.8512,
+      "step": 2262
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971868749429776e-05,
+      "loss": 16.6963,
+      "step": 2263
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971821134364687e-05,
+      "loss": 19.7512,
+      "step": 2264
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971773479150814e-05,
+      "loss": 17.571,
+      "step": 2265
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971725783788541e-05,
+      "loss": 18.418,
+      "step": 2266
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971678048278256e-05,
+      "loss": 17.3515,
+      "step": 2267
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971630272620343e-05,
+      "loss": 18.5153,
+      "step": 2268
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971582456815187e-05,
+      "loss": 15.7754,
+      "step": 2269
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971534600863174e-05,
+      "loss": 17.0,
+      "step": 2270
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971486704764692e-05,
+      "loss": 18.7497,
+      "step": 2271
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971438768520127e-05,
+      "loss": 17.9049,
+      "step": 2272
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971390792129866e-05,
+      "loss": 17.5415,
+      "step": 2273
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971342775594294e-05,
+      "loss": 15.3477,
+      "step": 2274
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971294718913803e-05,
+      "loss": 18.0271,
+      "step": 2275
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971246622088781e-05,
+      "loss": 15.9017,
+      "step": 2276
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971198485119613e-05,
+      "loss": 17.4437,
+      "step": 2277
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.97115030800669e-05,
+      "loss": 18.5255,
+      "step": 2278
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971102090750398e-05,
+      "loss": 16.5127,
+      "step": 2279
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.97105383335113e-05,
+      "loss": 17.141,
+      "step": 2280
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.971005535809275e-05,
+      "loss": 19.1885,
+      "step": 2281
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970957198125223e-05,
+      "loss": 17.6957,
+      "step": 2282
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970908820299363e-05,
+      "loss": 15.9941,
+      "step": 2283
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970860402332085e-05,
+      "loss": 18.3252,
+      "step": 2284
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970811944223784e-05,
+      "loss": 17.1584,
+      "step": 2285
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970763445974846e-05,
+      "loss": 18.3503,
+      "step": 2286
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970714907585668e-05,
+      "loss": 15.7732,
+      "step": 2287
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970666329056639e-05,
+      "loss": 16.928,
+      "step": 2288
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.97061771038815e-05,
+      "loss": 16.986,
+      "step": 2289
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970569051580596e-05,
+      "loss": 17.669,
+      "step": 2290
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.97052035263437e-05,
+      "loss": 16.8608,
+      "step": 2291
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970471613549863e-05,
+      "loss": 17.0007,
+      "step": 2292
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970422834327471e-05,
+      "loss": 17.9397,
+      "step": 2293
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970374014967587e-05,
+      "loss": 17.9308,
+      "step": 2294
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970325155470605e-05,
+      "loss": 15.1831,
+      "step": 2295
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.97027625583692e-05,
+      "loss": 18.8544,
+      "step": 2296
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970227316066926e-05,
+      "loss": 16.0644,
+      "step": 2297
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970178336161018e-05,
+      "loss": 15.2787,
+      "step": 2298
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970129316119593e-05,
+      "loss": 22.2124,
+      "step": 2299
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970080255943048e-05,
+      "loss": 17.9898,
+      "step": 2300
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.970031155631774e-05,
+      "loss": 17.074,
+      "step": 2301
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969982015186173e-05,
+      "loss": 17.2581,
+      "step": 2302
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.96993283460664e-05,
+      "loss": 17.7887,
+      "step": 2303
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.96988361389357e-05,
+      "loss": 17.6476,
+      "step": 2304
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969834353047363e-05,
+      "loss": 17.7689,
+      "step": 2305
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969785052068417e-05,
+      "loss": 17.8628,
+      "step": 2306
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969735710957129e-05,
+      "loss": 17.2488,
+      "step": 2307
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969686329713896e-05,
+      "loss": 18.7899,
+      "step": 2308
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.96963690833912e-05,
+      "loss": 17.2462,
+      "step": 2309
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969587446833197e-05,
+      "loss": 18.7532,
+      "step": 2310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969537945196529e-05,
+      "loss": 17.3876,
+      "step": 2311
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969488403429515e-05,
+      "loss": 19.2041,
+      "step": 2312
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969438821532554e-05,
+      "loss": 16.9453,
+      "step": 2313
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969389199506048e-05,
+      "loss": 18.6335,
+      "step": 2314
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969339537350395e-05,
+      "loss": 17.8587,
+      "step": 2315
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969289835066e-05,
+      "loss": 15.4424,
+      "step": 2316
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969240092653261e-05,
+      "loss": 18.3411,
+      "step": 2317
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969190310112579e-05,
+      "loss": 17.1016,
+      "step": 2318
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.96914048744436e-05,
+      "loss": 18.0147,
+      "step": 2319
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969090624649004e-05,
+      "loss": 19.6897,
+      "step": 2320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.969040721726914e-05,
+      "loss": 17.3208,
+      "step": 2321
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968990778678493e-05,
+      "loss": 18.8813,
+      "step": 2322
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968940795504145e-05,
+      "loss": 14.7969,
+      "step": 2323
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968890772204271e-05,
+      "loss": 18.417,
+      "step": 2324
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968840708779277e-05,
+      "loss": 18.7646,
+      "step": 2325
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968790605229567e-05,
+      "loss": 17.1894,
+      "step": 2326
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968740461555545e-05,
+      "loss": 17.116,
+      "step": 2327
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968690277757619e-05,
+      "loss": 17.7194,
+      "step": 2328
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968640053836188e-05,
+      "loss": 18.8914,
+      "step": 2329
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968589789791664e-05,
+      "loss": 18.2725,
+      "step": 2330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968539485624451e-05,
+      "loss": 18.1406,
+      "step": 2331
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968489141334954e-05,
+      "loss": 16.6925,
+      "step": 2332
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968438756923578e-05,
+      "loss": 17.5445,
+      "step": 2333
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968388332390733e-05,
+      "loss": 16.4495,
+      "step": 2334
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968337867736826e-05,
+      "loss": 15.7206,
+      "step": 2335
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968287362962264e-05,
+      "loss": 16.9881,
+      "step": 2336
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968236818067452e-05,
+      "loss": 16.7089,
+      "step": 2337
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968186233052801e-05,
+      "loss": 17.3465,
+      "step": 2338
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968135607918721e-05,
+      "loss": 17.6923,
+      "step": 2339
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968084942665616e-05,
+      "loss": 17.5411,
+      "step": 2340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.968034237293901e-05,
+      "loss": 16.6611,
+      "step": 2341
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.967983491803981e-05,
+      "loss": 17.6468,
+      "step": 2342
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967932706196267e-05,
+      "loss": 16.6493,
+      "step": 2343
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96788188047117e-05,
+      "loss": 15.6574,
+      "step": 2344
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967831014629099e-05,
+      "loss": 17.2943,
+      "step": 2345
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967780108670466e-05,
+      "loss": 17.3753,
+      "step": 2346
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967729162595682e-05,
+      "loss": 16.0548,
+      "step": 2347
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967678176405158e-05,
+      "loss": 18.1927,
+      "step": 2348
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967627150099306e-05,
+      "loss": 18.0261,
+      "step": 2349
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967576083678538e-05,
+      "loss": 17.7454,
+      "step": 2350
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967524977143265e-05,
+      "loss": 18.4236,
+      "step": 2351
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.9674738304939e-05,
+      "loss": 16.0963,
+      "step": 2352
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96742264373086e-05,
+      "loss": 16.2587,
+      "step": 2353
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967371416854553e-05,
+      "loss": 17.1081,
+      "step": 2354
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967320149865395e-05,
+      "loss": 20.1437,
+      "step": 2355
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.9672688427638e-05,
+      "loss": 16.761,
+      "step": 2356
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967217495550183e-05,
+      "loss": 16.8484,
+      "step": 2357
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967166108224957e-05,
+      "loss": 19.3989,
+      "step": 2358
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967114680788539e-05,
+      "loss": 15.2987,
+      "step": 2359
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967063213241343e-05,
+      "loss": 17.5897,
+      "step": 2360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.967011705583784e-05,
+      "loss": 15.1681,
+      "step": 2361
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.966960157816278e-05,
+      "loss": 17.2773,
+      "step": 2362
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.966908569939244e-05,
+      "loss": 17.969,
+      "step": 2363
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.966856941953096e-05,
+      "loss": 18.1553,
+      "step": 2364
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.966805273858251e-05,
+      "loss": 16.4788,
+      "step": 2365
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.966753565655127e-05,
+      "loss": 17.263,
+      "step": 2366
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96670181734414e-05,
+      "loss": 19.1864,
+      "step": 2367
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96665002892571e-05,
+      "loss": 17.1196,
+      "step": 2368
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.966598200400254e-05,
+      "loss": 16.3224,
+      "step": 2369
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.966546331768191e-05,
+      "loss": 15.633,
+      "step": 2370
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.966494423029939e-05,
+      "loss": 16.503,
+      "step": 2371
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96644247418592e-05,
+      "loss": 18.0755,
+      "step": 2372
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.966390485236549e-05,
+      "loss": 16.618,
+      "step": 2373
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96633845618225e-05,
+      "loss": 16.7189,
+      "step": 2374
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96628638702344e-05,
+      "loss": 16.6116,
+      "step": 2375
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.966234277760542e-05,
+      "loss": 18.3458,
+      "step": 2376
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.966182128393975e-05,
+      "loss": 17.2802,
+      "step": 2377
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.966129938924161e-05,
+      "loss": 17.6248,
+      "step": 2378
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.966077709351522e-05,
+      "loss": 18.4987,
+      "step": 2379
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.966025439676478e-05,
+      "loss": 18.5645,
+      "step": 2380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965973129899452e-05,
+      "loss": 16.8546,
+      "step": 2381
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965920780020867e-05,
+      "loss": 18.9657,
+      "step": 2382
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965868390041144e-05,
+      "loss": 18.3441,
+      "step": 2383
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965815959960708e-05,
+      "loss": 15.908,
+      "step": 2384
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965763489779982e-05,
+      "loss": 16.5176,
+      "step": 2385
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965710979499388e-05,
+      "loss": 17.2481,
+      "step": 2386
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965658429119354e-05,
+      "loss": 17.3833,
+      "step": 2387
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.9656058386403e-05,
+      "loss": 16.5865,
+      "step": 2388
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965553208062653e-05,
+      "loss": 16.8379,
+      "step": 2389
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965500537386836e-05,
+      "loss": 16.0869,
+      "step": 2390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965447826613277e-05,
+      "loss": 17.5114,
+      "step": 2391
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.9653950757424e-05,
+      "loss": 19.3513,
+      "step": 2392
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965342284774632e-05,
+      "loss": 18.0521,
+      "step": 2393
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965289453710398e-05,
+      "loss": 17.3319,
+      "step": 2394
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965236582550125e-05,
+      "loss": 17.2518,
+      "step": 2395
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965183671294242e-05,
+      "loss": 15.707,
+      "step": 2396
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965130719943174e-05,
+      "loss": 17.7864,
+      "step": 2397
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965077728497348e-05,
+      "loss": 17.7218,
+      "step": 2398
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.965024696957195e-05,
+      "loss": 17.9748,
+      "step": 2399
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96497162532314e-05,
+      "loss": 15.5512,
+      "step": 2400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964918513595612e-05,
+      "loss": 16.09,
+      "step": 2401
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964865361775042e-05,
+      "loss": 15.6861,
+      "step": 2402
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964812169861858e-05,
+      "loss": 17.6406,
+      "step": 2403
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964758937856489e-05,
+      "loss": 16.9158,
+      "step": 2404
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964705665759366e-05,
+      "loss": 17.3142,
+      "step": 2405
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964652353570918e-05,
+      "loss": 16.6811,
+      "step": 2406
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964599001291576e-05,
+      "loss": 18.1684,
+      "step": 2407
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964545608921772e-05,
+      "loss": 17.5541,
+      "step": 2408
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964492176461934e-05,
+      "loss": 17.1298,
+      "step": 2409
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964438703912498e-05,
+      "loss": 18.0053,
+      "step": 2410
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964385191273893e-05,
+      "loss": 16.2462,
+      "step": 2411
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96433163854655e-05,
+      "loss": 18.4111,
+      "step": 2412
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964278045730904e-05,
+      "loss": 17.5272,
+      "step": 2413
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964224412827385e-05,
+      "loss": 17.0879,
+      "step": 2414
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964170739836431e-05,
+      "loss": 16.8088,
+      "step": 2415
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96411702675847e-05,
+      "loss": 16.0326,
+      "step": 2416
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964063273593938e-05,
+      "loss": 17.7199,
+      "step": 2417
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.964009480343271e-05,
+      "loss": 17.8375,
+      "step": 2418
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.9639556470069e-05,
+      "loss": 16.9505,
+      "step": 2419
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.963901773585261e-05,
+      "loss": 16.8221,
+      "step": 2420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.963847860078788e-05,
+      "loss": 18.0265,
+      "step": 2421
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96379390648792e-05,
+      "loss": 17.3174,
+      "step": 2422
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.963739912813088e-05,
+      "loss": 17.2832,
+      "step": 2423
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.963685879054732e-05,
+      "loss": 14.9337,
+      "step": 2424
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.963631805213287e-05,
+      "loss": 16.179,
+      "step": 2425
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.963577691289188e-05,
+      "loss": 17.8611,
+      "step": 2426
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.963523537282873e-05,
+      "loss": 17.1458,
+      "step": 2427
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.963469343194781e-05,
+      "loss": 17.6714,
+      "step": 2428
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.963415109025347e-05,
+      "loss": 17.5464,
+      "step": 2429
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96336083477501e-05,
+      "loss": 19.4334,
+      "step": 2430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96330652044421e-05,
+      "loss": 16.608,
+      "step": 2431
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.963252166033384e-05,
+      "loss": 17.6649,
+      "step": 2432
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.963197771542969e-05,
+      "loss": 16.8486,
+      "step": 2433
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.963143336973408e-05,
+      "loss": 18.7233,
+      "step": 2434
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.963088862325138e-05,
+      "loss": 18.463,
+      "step": 2435
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.963034347598601e-05,
+      "loss": 19.211,
+      "step": 2436
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962979792794234e-05,
+      "loss": 18.9831,
+      "step": 2437
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96292519791248e-05,
+      "loss": 16.3438,
+      "step": 2438
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962870562953781e-05,
+      "loss": 17.0326,
+      "step": 2439
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962815887918576e-05,
+      "loss": 17.3197,
+      "step": 2440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962761172807305e-05,
+      "loss": 16.4956,
+      "step": 2441
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962706417620413e-05,
+      "loss": 16.9792,
+      "step": 2442
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962651622358341e-05,
+      "loss": 16.7389,
+      "step": 2443
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962596787021533e-05,
+      "loss": 16.2164,
+      "step": 2444
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962541911610429e-05,
+      "loss": 17.1686,
+      "step": 2445
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962486996125472e-05,
+      "loss": 17.2877,
+      "step": 2446
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962432040567108e-05,
+      "loss": 15.8329,
+      "step": 2447
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96237704493578e-05,
+      "loss": 17.3359,
+      "step": 2448
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962322009231932e-05,
+      "loss": 16.9153,
+      "step": 2449
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962266933456008e-05,
+      "loss": 14.9143,
+      "step": 2450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962211817608453e-05,
+      "loss": 16.2112,
+      "step": 2451
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962156661689712e-05,
+      "loss": 19.1049,
+      "step": 2452
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962101465700231e-05,
+      "loss": 14.9109,
+      "step": 2453
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.962046229640455e-05,
+      "loss": 17.8767,
+      "step": 2454
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96199095351083e-05,
+      "loss": 18.8924,
+      "step": 2455
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.961935637311804e-05,
+      "loss": 16.1863,
+      "step": 2456
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96188028104382e-05,
+      "loss": 18.884,
+      "step": 2457
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.961824884707329e-05,
+      "loss": 17.2703,
+      "step": 2458
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.961769448302774e-05,
+      "loss": 17.3887,
+      "step": 2459
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.961713971830609e-05,
+      "loss": 17.1053,
+      "step": 2460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.961658455291276e-05,
+      "loss": 18.2969,
+      "step": 2461
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.961602898685226e-05,
+      "loss": 18.1495,
+      "step": 2462
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.961547302012908e-05,
+      "loss": 16.2446,
+      "step": 2463
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.96149166527477e-05,
+      "loss": 16.5533,
+      "step": 2464
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.961435988471259e-05,
+      "loss": 16.9606,
+      "step": 2465
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.96138027160283e-05,
+      "loss": 16.3053,
+      "step": 2466
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.96132451466993e-05,
+      "loss": 17.7662,
+      "step": 2467
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.961268717673008e-05,
+      "loss": 15.3739,
+      "step": 2468
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.961212880612516e-05,
+      "loss": 18.7295,
+      "step": 2469
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.961157003488906e-05,
+      "loss": 17.695,
+      "step": 2470
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.961101086302626e-05,
+      "loss": 15.2801,
+      "step": 2471
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.961045129054132e-05,
+      "loss": 17.7201,
+      "step": 2472
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.960989131743871e-05,
+      "loss": 16.452,
+      "step": 2473
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.960933094372298e-05,
+      "loss": 18.7,
+      "step": 2474
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.960877016939867e-05,
+      "loss": 17.6527,
+      "step": 2475
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.960820899447027e-05,
+      "loss": 17.4913,
+      "step": 2476
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.960764741894233e-05,
+      "loss": 17.8108,
+      "step": 2477
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.96070854428194e-05,
+      "loss": 18.3287,
+      "step": 2478
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.9606523066106e-05,
+      "loss": 16.2727,
+      "step": 2479
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.960596028880666e-05,
+      "loss": 17.4541,
+      "step": 2480
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.960539711092596e-05,
+      "loss": 18.6549,
+      "step": 2481
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.960483353246842e-05,
+      "loss": 15.5347,
+      "step": 2482
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.96042695534386e-05,
+      "loss": 18.6649,
+      "step": 2483
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.960370517384106e-05,
+      "loss": 17.7745,
+      "step": 2484
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.960314039368033e-05,
+      "loss": 16.4767,
+      "step": 2485
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.960257521296102e-05,
+      "loss": 18.6946,
+      "step": 2486
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.960200963168765e-05,
+      "loss": 16.5317,
+      "step": 2487
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.96014436498648e-05,
+      "loss": 17.1018,
+      "step": 2488
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.960087726749703e-05,
+      "loss": 18.7494,
+      "step": 2489
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.960031048458895e-05,
+      "loss": 16.7651,
+      "step": 2490
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.95997433011451e-05,
+      "loss": 17.3635,
+      "step": 2491
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.959917571717008e-05,
+      "loss": 19.3554,
+      "step": 2492
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.959860773266847e-05,
+      "loss": 18.0366,
+      "step": 2493
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.959803934764484e-05,
+      "loss": 18.5308,
+      "step": 2494
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.95974705621038e-05,
+      "loss": 17.9706,
+      "step": 2495
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.959690137604994e-05,
+      "loss": 18.5054,
+      "step": 2496
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.959633178948787e-05,
+      "loss": 17.224,
+      "step": 2497
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.959576180242214e-05,
+      "loss": 17.93,
+      "step": 2498
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.95951914148574e-05,
+      "loss": 16.334,
+      "step": 2499
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.959462062679825e-05,
+      "loss": 16.0293,
+      "step": 2500
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.959404943824928e-05,
+      "loss": 17.696,
+      "step": 2501
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.95934778492151e-05,
+      "loss": 17.577,
+      "step": 2502
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.959290585970038e-05,
+      "loss": 15.0573,
+      "step": 2503
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.959233346970967e-05,
+      "loss": 16.0969,
+      "step": 2504
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.959176067924762e-05,
+      "loss": 19.7578,
+      "step": 2505
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.959118748831886e-05,
+      "loss": 17.7258,
+      "step": 2506
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.9590613896928e-05,
+      "loss": 18.0438,
+      "step": 2507
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.959003990507972e-05,
+      "loss": 17.7041,
+      "step": 2508
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958946551277859e-05,
+      "loss": 18.4223,
+      "step": 2509
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958889072002929e-05,
+      "loss": 19.8593,
+      "step": 2510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958831552683645e-05,
+      "loss": 17.6808,
+      "step": 2511
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958773993320472e-05,
+      "loss": 14.7581,
+      "step": 2512
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958716393913873e-05,
+      "loss": 16.9993,
+      "step": 2513
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958658754464318e-05,
+      "loss": 16.0945,
+      "step": 2514
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958601074972266e-05,
+      "loss": 15.3341,
+      "step": 2515
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958543355438187e-05,
+      "loss": 16.3138,
+      "step": 2516
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958485595862546e-05,
+      "loss": 17.9207,
+      "step": 2517
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958427796245808e-05,
+      "loss": 18.0252,
+      "step": 2518
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958369956588441e-05,
+      "loss": 17.1921,
+      "step": 2519
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958312076890913e-05,
+      "loss": 15.1345,
+      "step": 2520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958254157153691e-05,
+      "loss": 19.2492,
+      "step": 2521
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.95819619737724e-05,
+      "loss": 17.1148,
+      "step": 2522
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958138197562033e-05,
+      "loss": 15.4254,
+      "step": 2523
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958080157708535e-05,
+      "loss": 16.5303,
+      "step": 2524
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.958022077817216e-05,
+      "loss": 16.2234,
+      "step": 2525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957963957888542e-05,
+      "loss": 17.6389,
+      "step": 2526
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957905797922986e-05,
+      "loss": 18.2239,
+      "step": 2527
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957847597921015e-05,
+      "loss": 17.2669,
+      "step": 2528
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957789357883103e-05,
+      "loss": 19.3027,
+      "step": 2529
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957731077809716e-05,
+      "loss": 16.9578,
+      "step": 2530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957672757701327e-05,
+      "loss": 16.6611,
+      "step": 2531
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957614397558403e-05,
+      "loss": 17.6175,
+      "step": 2532
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957555997381422e-05,
+      "loss": 16.5339,
+      "step": 2533
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957497557170852e-05,
+      "loss": 16.9053,
+      "step": 2534
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957439076927164e-05,
+      "loss": 14.9791,
+      "step": 2535
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.95738055665083e-05,
+      "loss": 16.6148,
+      "step": 2536
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957321996342325e-05,
+      "loss": 16.1893,
+      "step": 2537
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957263396002121e-05,
+      "loss": 18.4072,
+      "step": 2538
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957204755630692e-05,
+      "loss": 18.0838,
+      "step": 2539
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957146075228508e-05,
+      "loss": 16.5415,
+      "step": 2540
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957087354796046e-05,
+      "loss": 17.1098,
+      "step": 2541
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.95702859433378e-05,
+      "loss": 15.6466,
+      "step": 2542
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.956969793842184e-05,
+      "loss": 17.4003,
+      "step": 2543
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.956910953321732e-05,
+      "loss": 17.1496,
+      "step": 2544
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.956852072772901e-05,
+      "loss": 15.3311,
+      "step": 2545
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.956793152196165e-05,
+      "loss": 17.7841,
+      "step": 2546
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.956734191592e-05,
+      "loss": 16.3827,
+      "step": 2547
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.956675190960882e-05,
+      "loss": 16.5722,
+      "step": 2548
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.95661615030329e-05,
+      "loss": 16.9388,
+      "step": 2549
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.956557069619696e-05,
+      "loss": 16.3282,
+      "step": 2550
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.956497948910581e-05,
+      "loss": 16.359,
+      "step": 2551
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.956438788176421e-05,
+      "loss": 16.8345,
+      "step": 2552
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.956379587417693e-05,
+      "loss": 16.3536,
+      "step": 2553
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.956320346634876e-05,
+      "loss": 18.2822,
+      "step": 2554
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.95626106582845e-05,
+      "loss": 17.8327,
+      "step": 2555
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.95620174499889e-05,
+      "loss": 18.9774,
+      "step": 2556
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.956142384146678e-05,
+      "loss": 18.4177,
+      "step": 2557
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.956082983272293e-05,
+      "loss": 18.1114,
+      "step": 2558
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.956023542376211e-05,
+      "loss": 16.8155,
+      "step": 2559
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955964061458918e-05,
+      "loss": 15.2453,
+      "step": 2560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.95590454052089e-05,
+      "loss": 16.7777,
+      "step": 2561
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955844979562609e-05,
+      "loss": 17.6282,
+      "step": 2562
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955785378584557e-05,
+      "loss": 17.6043,
+      "step": 2563
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955725737587214e-05,
+      "loss": 19.701,
+      "step": 2564
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955666056571063e-05,
+      "loss": 17.7881,
+      "step": 2565
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955606335536583e-05,
+      "loss": 18.6321,
+      "step": 2566
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955546574484259e-05,
+      "loss": 19.1156,
+      "step": 2567
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.95548677341457e-05,
+      "loss": 16.8465,
+      "step": 2568
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955426932328004e-05,
+      "loss": 17.5004,
+      "step": 2569
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955367051225043e-05,
+      "loss": 16.6647,
+      "step": 2570
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955307130106168e-05,
+      "loss": 16.3405,
+      "step": 2571
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955247168971864e-05,
+      "loss": 17.765,
+      "step": 2572
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955187167822616e-05,
+      "loss": 16.1296,
+      "step": 2573
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955127126658908e-05,
+      "loss": 16.6947,
+      "step": 2574
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955067045481224e-05,
+      "loss": 19.2787,
+      "step": 2575
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.955006924290051e-05,
+      "loss": 18.8649,
+      "step": 2576
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.954946763085872e-05,
+      "loss": 17.4993,
+      "step": 2577
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.954886561869177e-05,
+      "loss": 15.2315,
+      "step": 2578
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.954826320640448e-05,
+      "loss": 17.9094,
+      "step": 2579
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.954766039400173e-05,
+      "loss": 17.8238,
+      "step": 2580
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.954705718148837e-05,
+      "loss": 16.399,
+      "step": 2581
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.954645356886931e-05,
+      "loss": 17.7065,
+      "step": 2582
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.954584955614941e-05,
+      "loss": 14.8491,
+      "step": 2583
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.954524514333353e-05,
+      "loss": 18.0397,
+      "step": 2584
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.954464033042655e-05,
+      "loss": 18.2714,
+      "step": 2585
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.954403511743337e-05,
+      "loss": 16.9167,
+      "step": 2586
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.954342950435888e-05,
+      "loss": 17.0656,
+      "step": 2587
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.954282349120795e-05,
+      "loss": 20.3066,
+      "step": 2588
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.95422170779855e-05,
+      "loss": 16.414,
+      "step": 2589
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.95416102646964e-05,
+      "loss": 17.1019,
+      "step": 2590
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.954100305134557e-05,
+      "loss": 15.9901,
+      "step": 2591
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.954039543793791e-05,
+      "loss": 17.4528,
+      "step": 2592
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953978742447833e-05,
+      "loss": 16.612,
+      "step": 2593
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953917901097172e-05,
+      "loss": 15.7151,
+      "step": 2594
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.9538570197423e-05,
+      "loss": 17.674,
+      "step": 2595
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953796098383712e-05,
+      "loss": 17.0736,
+      "step": 2596
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953735137021895e-05,
+      "loss": 16.6013,
+      "step": 2597
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953674135657345e-05,
+      "loss": 17.588,
+      "step": 2598
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953613094290553e-05,
+      "loss": 19.5076,
+      "step": 2599
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953552012922012e-05,
+      "loss": 17.6071,
+      "step": 2600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953490891552212e-05,
+      "loss": 16.7973,
+      "step": 2601
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953429730181653e-05,
+      "loss": 16.8934,
+      "step": 2602
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953368528810827e-05,
+      "loss": 15.6101,
+      "step": 2603
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953307287440226e-05,
+      "loss": 17.1223,
+      "step": 2604
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953246006070344e-05,
+      "loss": 16.1726,
+      "step": 2605
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953184684701678e-05,
+      "loss": 18.1126,
+      "step": 2606
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953123323334723e-05,
+      "loss": 15.4048,
+      "step": 2607
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953061921969974e-05,
+      "loss": 16.9521,
+      "step": 2608
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.953000480607927e-05,
+      "loss": 15.4494,
+      "step": 2609
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952938999249078e-05,
+      "loss": 15.523,
+      "step": 2610
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952877477893922e-05,
+      "loss": 18.9187,
+      "step": 2611
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.95281591654296e-05,
+      "loss": 18.6421,
+      "step": 2612
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952754315196687e-05,
+      "loss": 16.8784,
+      "step": 2613
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952692673855597e-05,
+      "loss": 17.9069,
+      "step": 2614
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952630992520191e-05,
+      "loss": 18.9626,
+      "step": 2615
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952569271190968e-05,
+      "loss": 16.0564,
+      "step": 2616
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952507509868425e-05,
+      "loss": 15.24,
+      "step": 2617
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952445708553062e-05,
+      "loss": 15.6846,
+      "step": 2618
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952383867245375e-05,
+      "loss": 19.1895,
+      "step": 2619
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952321985945867e-05,
+      "loss": 15.8497,
+      "step": 2620
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952260064655035e-05,
+      "loss": 17.7723,
+      "step": 2621
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952198103373381e-05,
+      "loss": 18.4264,
+      "step": 2622
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952136102101404e-05,
+      "loss": 17.7423,
+      "step": 2623
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952074060839604e-05,
+      "loss": 18.0445,
+      "step": 2624
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952011979588485e-05,
+      "loss": 16.9507,
+      "step": 2625
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951949858348546e-05,
+      "loss": 17.1322,
+      "step": 2626
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951887697120288e-05,
+      "loss": 20.2021,
+      "step": 2627
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951825495904215e-05,
+      "loss": 17.6453,
+      "step": 2628
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951763254700829e-05,
+      "loss": 15.5519,
+      "step": 2629
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951700973510632e-05,
+      "loss": 16.9134,
+      "step": 2630
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951638652334125e-05,
+      "loss": 18.1778,
+      "step": 2631
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951576291171816e-05,
+      "loss": 17.3184,
+      "step": 2632
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951513890024203e-05,
+      "loss": 16.8402,
+      "step": 2633
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951451448891795e-05,
+      "loss": 18.5882,
+      "step": 2634
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951388967775093e-05,
+      "loss": 17.6842,
+      "step": 2635
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951326446674602e-05,
+      "loss": 17.0152,
+      "step": 2636
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951263885590828e-05,
+      "loss": 19.0478,
+      "step": 2637
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951201284524275e-05,
+      "loss": 17.2179,
+      "step": 2638
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951138643475449e-05,
+      "loss": 14.7412,
+      "step": 2639
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951075962444856e-05,
+      "loss": 17.5971,
+      "step": 2640
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.951013241433003e-05,
+      "loss": 15.7081,
+      "step": 2641
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950950480440395e-05,
+      "loss": 17.9462,
+      "step": 2642
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.95088767946754e-05,
+      "loss": 14.9061,
+      "step": 2643
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950824838514943e-05,
+      "loss": 19.5591,
+      "step": 2644
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950761957583115e-05,
+      "loss": 17.3762,
+      "step": 2645
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950699036672559e-05,
+      "loss": 17.3616,
+      "step": 2646
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950636075783787e-05,
+      "loss": 17.9629,
+      "step": 2647
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950573074917306e-05,
+      "loss": 16.5707,
+      "step": 2648
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950510034073627e-05,
+      "loss": 16.4965,
+      "step": 2649
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950446953253257e-05,
+      "loss": 16.8055,
+      "step": 2650
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950383832456702e-05,
+      "loss": 20.5091,
+      "step": 2651
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950320671684479e-05,
+      "loss": 16.1638,
+      "step": 2652
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950257470937092e-05,
+      "loss": 17.0265,
+      "step": 2653
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950194230215054e-05,
+      "loss": 17.0472,
+      "step": 2654
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950130949518876e-05,
+      "loss": 19.8464,
+      "step": 2655
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950067628849068e-05,
+      "loss": 18.0344,
+      "step": 2656
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950004268206141e-05,
+      "loss": 20.1587,
+      "step": 2657
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.949940867590608e-05,
+      "loss": 18.6745,
+      "step": 2658
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.94987742700298e-05,
+      "loss": 18.5714,
+      "step": 2659
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.94981394644377e-05,
+      "loss": 18.3475,
+      "step": 2660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.94975042591349e-05,
+      "loss": 17.1919,
+      "step": 2661
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.949686865412653e-05,
+      "loss": 18.2599,
+      "step": 2662
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.949623264941771e-05,
+      "loss": 16.6569,
+      "step": 2663
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.949559624501362e-05,
+      "loss": 17.0588,
+      "step": 2664
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.949495944091935e-05,
+      "loss": 19.0448,
+      "step": 2665
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.949432223714005e-05,
+      "loss": 17.5271,
+      "step": 2666
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.94936846336809e-05,
+      "loss": 17.1431,
+      "step": 2667
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.949304663054702e-05,
+      "loss": 17.3941,
+      "step": 2668
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.949240822774358e-05,
+      "loss": 19.0199,
+      "step": 2669
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.949176942527569e-05,
+      "loss": 18.2656,
+      "step": 2670
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.949113022314855e-05,
+      "loss": 16.8852,
+      "step": 2671
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.949049062136733e-05,
+      "loss": 17.0725,
+      "step": 2672
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.948985061993718e-05,
+      "loss": 15.5003,
+      "step": 2673
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.948921021886325e-05,
+      "loss": 19.3096,
+      "step": 2674
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.948856941815074e-05,
+      "loss": 17.1163,
+      "step": 2675
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.948792821780482e-05,
+      "loss": 16.7022,
+      "step": 2676
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.948728661783064e-05,
+      "loss": 18.2934,
+      "step": 2677
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.94866446182334e-05,
+      "loss": 17.8245,
+      "step": 2678
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.94860022190183e-05,
+      "loss": 17.8569,
+      "step": 2679
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.948535942019051e-05,
+      "loss": 16.0816,
+      "step": 2680
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.948471622175523e-05,
+      "loss": 17.9044,
+      "step": 2681
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.948407262371764e-05,
+      "loss": 16.411,
+      "step": 2682
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.948342862608295e-05,
+      "loss": 16.6305,
+      "step": 2683
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.948278422885635e-05,
+      "loss": 17.0858,
+      "step": 2684
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.948213943204305e-05,
+      "loss": 16.2829,
+      "step": 2685
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.948149423564825e-05,
+      "loss": 20.009,
+      "step": 2686
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.94808486396772e-05,
+      "loss": 16.3466,
+      "step": 2687
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.948020264413505e-05,
+      "loss": 17.8709,
+      "step": 2688
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.947955624902705e-05,
+      "loss": 16.7171,
+      "step": 2689
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.94789094543584e-05,
+      "loss": 17.3307,
+      "step": 2690
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.947826226013437e-05,
+      "loss": 16.4413,
+      "step": 2691
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.947761466636014e-05,
+      "loss": 17.7927,
+      "step": 2692
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.947696667304097e-05,
+      "loss": 17.9061,
+      "step": 2693
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.947631828018206e-05,
+      "loss": 15.7294,
+      "step": 2694
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.947566948778867e-05,
+      "loss": 20.1731,
+      "step": 2695
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.947502029586603e-05,
+      "loss": 15.0808,
+      "step": 2696
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.947437070441938e-05,
+      "loss": 14.4708,
+      "step": 2697
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.947372071345399e-05,
+      "loss": 16.6087,
+      "step": 2698
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.947307032297508e-05,
+      "loss": 16.4289,
+      "step": 2699
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.947241953298791e-05,
+      "loss": 19.1618,
+      "step": 2700
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.947176834349774e-05,
+      "loss": 18.3286,
+      "step": 2701
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.947111675450982e-05,
+      "loss": 17.7453,
+      "step": 2702
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.947046476602941e-05,
+      "loss": 18.979,
+      "step": 2703
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.946981237806181e-05,
+      "loss": 16.0462,
+      "step": 2704
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.946915959061224e-05,
+      "loss": 18.5312,
+      "step": 2705
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.9468506403686e-05,
+      "loss": 21.5092,
+      "step": 2706
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.946785281728835e-05,
+      "loss": 17.2237,
+      "step": 2707
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.946719883142458e-05,
+      "loss": 17.6826,
+      "step": 2708
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.946654444609997e-05,
+      "loss": 18.0743,
+      "step": 2709
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.946588966131978e-05,
+      "loss": 16.8756,
+      "step": 2710
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.946523447708933e-05,
+      "loss": 17.5173,
+      "step": 2711
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.946457889341391e-05,
+      "loss": 18.4133,
+      "step": 2712
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94639229102988e-05,
+      "loss": 18.267,
+      "step": 2713
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.946326652774929e-05,
+      "loss": 15.03,
+      "step": 2714
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.946260974577068e-05,
+      "loss": 16.2758,
+      "step": 2715
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94619525643683e-05,
+      "loss": 15.2372,
+      "step": 2716
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.946129498354744e-05,
+      "loss": 16.3028,
+      "step": 2717
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94606370033134e-05,
+      "loss": 16.3787,
+      "step": 2718
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.945997862367152e-05,
+      "loss": 18.2308,
+      "step": 2719
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.945931984462709e-05,
+      "loss": 15.9026,
+      "step": 2720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.945866066618544e-05,
+      "loss": 17.2072,
+      "step": 2721
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94580010883519e-05,
+      "loss": 15.827,
+      "step": 2722
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.945734111113178e-05,
+      "loss": 17.4664,
+      "step": 2723
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.945668073453042e-05,
+      "loss": 17.6792,
+      "step": 2724
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.945601995855318e-05,
+      "loss": 16.2016,
+      "step": 2725
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.945535878320534e-05,
+      "loss": 18.3041,
+      "step": 2726
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.945469720849227e-05,
+      "loss": 17.8752,
+      "step": 2727
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.945403523441932e-05,
+      "loss": 17.1607,
+      "step": 2728
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.945337286099182e-05,
+      "loss": 17.7405,
+      "step": 2729
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.945271008821513e-05,
+      "loss": 18.6995,
+      "step": 2730
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.945204691609458e-05,
+      "loss": 18.744,
+      "step": 2731
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.945138334463556e-05,
+      "loss": 17.6015,
+      "step": 2732
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94507193738434e-05,
+      "loss": 20.0197,
+      "step": 2733
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.945005500372346e-05,
+      "loss": 18.3303,
+      "step": 2734
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944939023428114e-05,
+      "loss": 17.7166,
+      "step": 2735
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944872506552177e-05,
+      "loss": 16.2915,
+      "step": 2736
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944805949745075e-05,
+      "loss": 19.8648,
+      "step": 2737
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944739353007344e-05,
+      "loss": 15.9789,
+      "step": 2738
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944672716339521e-05,
+      "loss": 17.3529,
+      "step": 2739
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944606039742144e-05,
+      "loss": 18.4172,
+      "step": 2740
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944539323215754e-05,
+      "loss": 16.2736,
+      "step": 2741
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944472566760887e-05,
+      "loss": 17.8151,
+      "step": 2742
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944405770378085e-05,
+      "loss": 15.6984,
+      "step": 2743
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944338934067884e-05,
+      "loss": 17.237,
+      "step": 2744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944272057830826e-05,
+      "loss": 19.4309,
+      "step": 2745
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944205141667451e-05,
+      "loss": 17.0662,
+      "step": 2746
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944138185578298e-05,
+      "loss": 16.9425,
+      "step": 2747
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944071189563909e-05,
+      "loss": 17.2255,
+      "step": 2748
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.944004153624825e-05,
+      "loss": 15.6723,
+      "step": 2749
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.943937077761585e-05,
+      "loss": 19.0832,
+      "step": 2750
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.943869961974733e-05,
+      "loss": 16.7487,
+      "step": 2751
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94380280626481e-05,
+      "loss": 16.1101,
+      "step": 2752
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94373561063236e-05,
+      "loss": 16.8622,
+      "step": 2753
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.943668375077925e-05,
+      "loss": 20.9694,
+      "step": 2754
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.943601099602046e-05,
+      "loss": 17.9478,
+      "step": 2755
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.943533784205267e-05,
+      "loss": 16.8595,
+      "step": 2756
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.943466428888135e-05,
+      "loss": 17.8177,
+      "step": 2757
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.943399033651189e-05,
+      "loss": 16.7378,
+      "step": 2758
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.943331598494973e-05,
+      "loss": 17.3122,
+      "step": 2759
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.943264123420037e-05,
+      "loss": 18.8788,
+      "step": 2760
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.943196608426924e-05,
+      "loss": 17.6248,
+      "step": 2761
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.943129053516176e-05,
+      "loss": 16.2226,
+      "step": 2762
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94306145868834e-05,
+      "loss": 20.196,
+      "step": 2763
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.942993823943962e-05,
+      "loss": 15.8753,
+      "step": 2764
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94292614928359e-05,
+      "loss": 17.4296,
+      "step": 2765
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94285843470777e-05,
+      "loss": 16.3373,
+      "step": 2766
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.942790680217044e-05,
+      "loss": 16.0668,
+      "step": 2767
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.942722885811965e-05,
+      "loss": 16.6853,
+      "step": 2768
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94265505149308e-05,
+      "loss": 17.747,
+      "step": 2769
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.942587177260933e-05,
+      "loss": 17.3111,
+      "step": 2770
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.942519263116074e-05,
+      "loss": 15.7881,
+      "step": 2771
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.942451309059054e-05,
+      "loss": 16.6374,
+      "step": 2772
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.942383315090418e-05,
+      "loss": 17.5407,
+      "step": 2773
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.942315281210717e-05,
+      "loss": 17.1656,
+      "step": 2774
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.9422472074205e-05,
+      "loss": 19.2975,
+      "step": 2775
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.942179093720317e-05,
+      "loss": 19.7152,
+      "step": 2776
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.942110940110716e-05,
+      "loss": 15.2487,
+      "step": 2777
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94204274659225e-05,
+      "loss": 19.0105,
+      "step": 2778
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.941974513165468e-05,
+      "loss": 16.2002,
+      "step": 2779
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.941906239830922e-05,
+      "loss": 18.0728,
+      "step": 2780
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.941837926589163e-05,
+      "loss": 15.8647,
+      "step": 2781
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.941769573440745e-05,
+      "loss": 17.7527,
+      "step": 2782
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.941701180386214e-05,
+      "loss": 18.3717,
+      "step": 2783
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.941632747426129e-05,
+      "loss": 17.0581,
+      "step": 2784
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.941564274561037e-05,
+      "loss": 18.6885,
+      "step": 2785
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.941495761791494e-05,
+      "loss": 16.7796,
+      "step": 2786
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.941427209118055e-05,
+      "loss": 17.7567,
+      "step": 2787
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94135861654127e-05,
+      "loss": 15.5161,
+      "step": 2788
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.941289984061694e-05,
+      "loss": 18.6221,
+      "step": 2789
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94122131167988e-05,
+      "loss": 17.9802,
+      "step": 2790
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.941152599396387e-05,
+      "loss": 18.1371,
+      "step": 2791
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.941083847211765e-05,
+      "loss": 20.0114,
+      "step": 2792
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.941015055126571e-05,
+      "loss": 18.3409,
+      "step": 2793
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.940946223141361e-05,
+      "loss": 15.0644,
+      "step": 2794
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94087735125669e-05,
+      "loss": 17.8009,
+      "step": 2795
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.940808439473116e-05,
+      "loss": 18.8587,
+      "step": 2796
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.940739487791194e-05,
+      "loss": 17.9312,
+      "step": 2797
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94067049621148e-05,
+      "loss": 16.5767,
+      "step": 2798
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.940601464734531e-05,
+      "loss": 15.9185,
+      "step": 2799
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.940532393360906e-05,
+      "loss": 17.5916,
+      "step": 2800
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.940463282091162e-05,
+      "loss": 17.9175,
+      "step": 2801
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.940394130925859e-05,
+      "loss": 18.3339,
+      "step": 2802
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.940324939865553e-05,
+      "loss": 17.2584,
+      "step": 2803
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.940255708910802e-05,
+      "loss": 17.6738,
+      "step": 2804
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.940186438062168e-05,
+      "loss": 18.041,
+      "step": 2805
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.940117127320207e-05,
+      "loss": 17.3155,
+      "step": 2806
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.940047776685484e-05,
+      "loss": 16.8595,
+      "step": 2807
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.939978386158553e-05,
+      "loss": 15.3598,
+      "step": 2808
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.939908955739979e-05,
+      "loss": 19.5491,
+      "step": 2809
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.939839485430318e-05,
+      "loss": 17.9738,
+      "step": 2810
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.939769975230135e-05,
+      "loss": 18.4333,
+      "step": 2811
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.939700425139989e-05,
+      "loss": 16.8055,
+      "step": 2812
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.939630835160444e-05,
+      "loss": 17.004,
+      "step": 2813
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.93956120529206e-05,
+      "loss": 17.0534,
+      "step": 2814
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.939491535535399e-05,
+      "loss": 17.4137,
+      "step": 2815
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.939421825891027e-05,
+      "loss": 15.945,
+      "step": 2816
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.939352076359501e-05,
+      "loss": 18.3421,
+      "step": 2817
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.93928228694139e-05,
+      "loss": 18.5471,
+      "step": 2818
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.939212457637253e-05,
+      "loss": 18.5774,
+      "step": 2819
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.939142588447658e-05,
+      "loss": 16.9787,
+      "step": 2820
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.939072679373167e-05,
+      "loss": 17.3306,
+      "step": 2821
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.939002730414344e-05,
+      "loss": 16.3047,
+      "step": 2822
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.938932741571755e-05,
+      "loss": 16.1709,
+      "step": 2823
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.938862712845964e-05,
+      "loss": 16.5021,
+      "step": 2824
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.938792644237537e-05,
+      "loss": 17.2099,
+      "step": 2825
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.938722535747043e-05,
+      "loss": 15.837,
+      "step": 2826
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.938652387375044e-05,
+      "loss": 17.9309,
+      "step": 2827
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.938582199122106e-05,
+      "loss": 18.2907,
+      "step": 2828
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.9385119709888e-05,
+      "loss": 17.422,
+      "step": 2829
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.938441702975689e-05,
+      "loss": 19.0927,
+      "step": 2830
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.938371395083344e-05,
+      "loss": 17.4605,
+      "step": 2831
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.938301047312328e-05,
+      "loss": 17.4722,
+      "step": 2832
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.938230659663213e-05,
+      "loss": 18.8042,
+      "step": 2833
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.938160232136568e-05,
+      "loss": 16.8977,
+      "step": 2834
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.938089764732958e-05,
+      "loss": 17.7622,
+      "step": 2835
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.938019257452955e-05,
+      "loss": 18.3183,
+      "step": 2836
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.937948710297127e-05,
+      "loss": 17.3179,
+      "step": 2837
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.937878123266044e-05,
+      "loss": 16.8317,
+      "step": 2838
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.937807496360277e-05,
+      "loss": 17.1613,
+      "step": 2839
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.937736829580395e-05,
+      "loss": 17.4948,
+      "step": 2840
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93766612292697e-05,
+      "loss": 18.3129,
+      "step": 2841
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93759537640057e-05,
+      "loss": 18.3498,
+      "step": 2842
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93752459000177e-05,
+      "loss": 18.707,
+      "step": 2843
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93745376373114e-05,
+      "loss": 18.4156,
+      "step": 2844
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.937382897589251e-05,
+      "loss": 14.9891,
+      "step": 2845
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.937311991576678e-05,
+      "loss": 16.3911,
+      "step": 2846
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93724104569399e-05,
+      "loss": 17.5761,
+      "step": 2847
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.937170059941763e-05,
+      "loss": 18.4755,
+      "step": 2848
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.937099034320568e-05,
+      "loss": 16.9648,
+      "step": 2849
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93702796883098e-05,
+      "loss": 17.6781,
+      "step": 2850
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.936956863473572e-05,
+      "loss": 18.0324,
+      "step": 2851
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.936885718248919e-05,
+      "loss": 17.0272,
+      "step": 2852
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.936814533157594e-05,
+      "loss": 17.285,
+      "step": 2853
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.936743308200176e-05,
+      "loss": 18.3752,
+      "step": 2854
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.936672043377234e-05,
+      "loss": 16.5879,
+      "step": 2855
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.936600738689348e-05,
+      "loss": 16.9276,
+      "step": 2856
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.936529394137091e-05,
+      "loss": 17.7555,
+      "step": 2857
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.936458009721041e-05,
+      "loss": 17.7086,
+      "step": 2858
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.936386585441775e-05,
+      "loss": 17.528,
+      "step": 2859
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.936315121299868e-05,
+      "loss": 18.6991,
+      "step": 2860
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.936243617295899e-05,
+      "loss": 16.3361,
+      "step": 2861
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.936172073430442e-05,
+      "loss": 17.6422,
+      "step": 2862
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93610048970408e-05,
+      "loss": 16.8885,
+      "step": 2863
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.936028866117385e-05,
+      "loss": 18.1554,
+      "step": 2864
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.935957202670937e-05,
+      "loss": 17.7792,
+      "step": 2865
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.935885499365321e-05,
+      "loss": 17.9492,
+      "step": 2866
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.935813756201108e-05,
+      "loss": 16.6578,
+      "step": 2867
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93574197317888e-05,
+      "loss": 16.5755,
+      "step": 2868
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.935670150299219e-05,
+      "loss": 18.1975,
+      "step": 2869
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.935598287562702e-05,
+      "loss": 17.781,
+      "step": 2870
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.935526384969911e-05,
+      "loss": 17.1698,
+      "step": 2871
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.935454442521426e-05,
+      "loss": 14.8855,
+      "step": 2872
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.935382460217828e-05,
+      "loss": 16.6533,
+      "step": 2873
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.935310438059699e-05,
+      "loss": 16.4397,
+      "step": 2874
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.935238376047618e-05,
+      "loss": 19.5977,
+      "step": 2875
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93516627418217e-05,
+      "loss": 16.6837,
+      "step": 2876
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.935094132463936e-05,
+      "loss": 18.6168,
+      "step": 2877
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.9350219508935e-05,
+      "loss": 18.4636,
+      "step": 2878
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.934949729471441e-05,
+      "loss": 17.0282,
+      "step": 2879
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.934877468198345e-05,
+      "loss": 17.7088,
+      "step": 2880
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.934805167074796e-05,
+      "loss": 17.9444,
+      "step": 2881
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.934732826101377e-05,
+      "loss": 16.7786,
+      "step": 2882
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.934660445278673e-05,
+      "loss": 16.4521,
+      "step": 2883
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.934588024607268e-05,
+      "loss": 17.6399,
+      "step": 2884
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.934515564087746e-05,
+      "loss": 18.7478,
+      "step": 2885
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.934443063720691e-05,
+      "loss": 17.6573,
+      "step": 2886
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93437052350669e-05,
+      "loss": 17.0956,
+      "step": 2887
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.934297943446333e-05,
+      "loss": 18.0833,
+      "step": 2888
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.934225323540198e-05,
+      "loss": 17.619,
+      "step": 2889
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.934152663788878e-05,
+      "loss": 18.35,
+      "step": 2890
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.934079964192956e-05,
+      "loss": 18.5931,
+      "step": 2891
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.934007224753022e-05,
+      "loss": 18.7299,
+      "step": 2892
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93393444546966e-05,
+      "loss": 16.328,
+      "step": 2893
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.933861626343459e-05,
+      "loss": 18.9917,
+      "step": 2894
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.933788767375007e-05,
+      "loss": 18.6824,
+      "step": 2895
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.933715868564895e-05,
+      "loss": 17.6403,
+      "step": 2896
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.933642929913708e-05,
+      "loss": 18.1262,
+      "step": 2897
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.933569951422038e-05,
+      "loss": 15.0746,
+      "step": 2898
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.933496933090471e-05,
+      "loss": 18.0775,
+      "step": 2899
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.933423874919599e-05,
+      "loss": 18.2938,
+      "step": 2900
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.933350776910012e-05,
+      "loss": 18.3471,
+      "step": 2901
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.933277639062298e-05,
+      "loss": 16.6627,
+      "step": 2902
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93320446137705e-05,
+      "loss": 17.5312,
+      "step": 2903
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.933131243854859e-05,
+      "loss": 15.6449,
+      "step": 2904
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.933057986496315e-05,
+      "loss": 16.6716,
+      "step": 2905
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93298468930201e-05,
+      "loss": 17.9206,
+      "step": 2906
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.932911352272536e-05,
+      "loss": 16.6557,
+      "step": 2907
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.932837975408485e-05,
+      "loss": 17.0572,
+      "step": 2908
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93276455871045e-05,
+      "loss": 16.4312,
+      "step": 2909
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.932691102179024e-05,
+      "loss": 16.9456,
+      "step": 2910
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.932617605814799e-05,
+      "loss": 16.6951,
+      "step": 2911
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93254406961837e-05,
+      "loss": 17.0905,
+      "step": 2912
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.932470493590328e-05,
+      "loss": 18.832,
+      "step": 2913
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.932396877731271e-05,
+      "loss": 17.9747,
+      "step": 2914
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.932323222041793e-05,
+      "loss": 16.82,
+      "step": 2915
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.932249526522486e-05,
+      "loss": 16.1074,
+      "step": 2916
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.932175791173945e-05,
+      "loss": 17.8149,
+      "step": 2917
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93210201599677e-05,
+      "loss": 17.5568,
+      "step": 2918
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93202820099155e-05,
+      "loss": 18.2299,
+      "step": 2919
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.931954346158888e-05,
+      "loss": 17.7639,
+      "step": 2920
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.931880451499375e-05,
+      "loss": 16.7066,
+      "step": 2921
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.931806517013612e-05,
+      "loss": 16.2384,
+      "step": 2922
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.931732542702193e-05,
+      "loss": 16.7717,
+      "step": 2923
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.931658528565718e-05,
+      "loss": 17.8412,
+      "step": 2924
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93158447460478e-05,
+      "loss": 17.6908,
+      "step": 2925
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.931510380819982e-05,
+      "loss": 17.023,
+      "step": 2926
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93143624721192e-05,
+      "loss": 16.9356,
+      "step": 2927
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.931362073781193e-05,
+      "loss": 18.8993,
+      "step": 2928
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.9312878605284e-05,
+      "loss": 18.1288,
+      "step": 2929
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.931213607454139e-05,
+      "loss": 18.2109,
+      "step": 2930
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.931139314559014e-05,
+      "loss": 16.9422,
+      "step": 2931
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93106498184362e-05,
+      "loss": 16.6724,
+      "step": 2932
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.930990609308559e-05,
+      "loss": 17.8611,
+      "step": 2933
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.930916196954431e-05,
+      "loss": 18.9505,
+      "step": 2934
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.930841744781838e-05,
+      "loss": 16.5875,
+      "step": 2935
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.930767252791381e-05,
+      "loss": 17.9521,
+      "step": 2936
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.930692720983662e-05,
+      "loss": 17.0746,
+      "step": 2937
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.93061814935928e-05,
+      "loss": 17.4713,
+      "step": 2938
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.930543537918843e-05,
+      "loss": 17.1172,
+      "step": 2939
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.930468886662947e-05,
+      "loss": 16.0495,
+      "step": 2940
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.930394195592199e-05,
+      "loss": 17.935,
+      "step": 2941
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.9303194647072e-05,
+      "loss": 16.9229,
+      "step": 2942
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.930244694008555e-05,
+      "loss": 18.8433,
+      "step": 2943
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.930169883496867e-05,
+      "loss": 16.7015,
+      "step": 2944
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.930095033172739e-05,
+      "loss": 16.4476,
+      "step": 2945
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.930020143036779e-05,
+      "loss": 16.0221,
+      "step": 2946
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.929945213089589e-05,
+      "loss": 17.2464,
+      "step": 2947
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.929870243331774e-05,
+      "loss": 16.1401,
+      "step": 2948
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92979523376394e-05,
+      "loss": 17.5806,
+      "step": 2949
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92972018438669e-05,
+      "loss": 15.4039,
+      "step": 2950
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.929645095200636e-05,
+      "loss": 18.0228,
+      "step": 2951
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92956996620638e-05,
+      "loss": 18.7383,
+      "step": 2952
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92949479740453e-05,
+      "loss": 16.3939,
+      "step": 2953
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.929419588795693e-05,
+      "loss": 18.6429,
+      "step": 2954
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.929344340380475e-05,
+      "loss": 18.0197,
+      "step": 2955
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.929269052159486e-05,
+      "loss": 19.0072,
+      "step": 2956
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.929193724133333e-05,
+      "loss": 17.7358,
+      "step": 2957
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.929118356302621e-05,
+      "loss": 16.644,
+      "step": 2958
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.929042948667964e-05,
+      "loss": 16.1216,
+      "step": 2959
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.928967501229968e-05,
+      "loss": 16.7025,
+      "step": 2960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.928892013989244e-05,
+      "loss": 15.756,
+      "step": 2961
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.928816486946399e-05,
+      "loss": 19.3912,
+      "step": 2962
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.928740920102044e-05,
+      "loss": 16.6582,
+      "step": 2963
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.928665313456791e-05,
+      "loss": 18.8337,
+      "step": 2964
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.928589667011248e-05,
+      "loss": 16.7363,
+      "step": 2965
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.928513980766026e-05,
+      "loss": 18.0168,
+      "step": 2966
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92843825472174e-05,
+      "loss": 17.4233,
+      "step": 2967
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.928362488878996e-05,
+      "loss": 16.953,
+      "step": 2968
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.928286683238409e-05,
+      "loss": 16.8351,
+      "step": 2969
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.928210837800591e-05,
+      "loss": 17.7427,
+      "step": 2970
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.928134952566153e-05,
+      "loss": 15.7689,
+      "step": 2971
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92805902753571e-05,
+      "loss": 17.1306,
+      "step": 2972
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.927983062709874e-05,
+      "loss": 17.0209,
+      "step": 2973
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.927907058089258e-05,
+      "loss": 16.7538,
+      "step": 2974
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.927831013674477e-05,
+      "loss": 17.3207,
+      "step": 2975
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.927754929466144e-05,
+      "loss": 17.0229,
+      "step": 2976
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.927678805464873e-05,
+      "loss": 19.4199,
+      "step": 2977
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.927602641671279e-05,
+      "loss": 17.4352,
+      "step": 2978
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.927526438085979e-05,
+      "loss": 18.3114,
+      "step": 2979
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.927450194709584e-05,
+      "loss": 16.7102,
+      "step": 2980
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.927373911542714e-05,
+      "loss": 17.425,
+      "step": 2981
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.927297588585984e-05,
+      "loss": 16.3013,
+      "step": 2982
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.927221225840011e-05,
+      "loss": 16.155,
+      "step": 2983
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.927144823305407e-05,
+      "loss": 18.865,
+      "step": 2984
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.927068380982795e-05,
+      "loss": 15.6463,
+      "step": 2985
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.926991898872789e-05,
+      "loss": 19.0828,
+      "step": 2986
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.926915376976007e-05,
+      "loss": 16.6514,
+      "step": 2987
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92683881529307e-05,
+      "loss": 17.3275,
+      "step": 2988
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.926762213824589e-05,
+      "loss": 16.5022,
+      "step": 2989
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.926685572571191e-05,
+      "loss": 18.0999,
+      "step": 2990
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.926608891533487e-05,
+      "loss": 18.3929,
+      "step": 2991
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.926532170712105e-05,
+      "loss": 17.0507,
+      "step": 2992
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.926455410107656e-05,
+      "loss": 16.7177,
+      "step": 2993
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.926378609720763e-05,
+      "loss": 17.6168,
+      "step": 2994
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.926301769552049e-05,
+      "loss": 16.9854,
+      "step": 2995
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.926224889602133e-05,
+      "loss": 17.6353,
+      "step": 2996
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.926147969871633e-05,
+      "loss": 16.4997,
+      "step": 2997
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.926071010361173e-05,
+      "loss": 16.77,
+      "step": 2998
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.925994011071374e-05,
+      "loss": 16.4505,
+      "step": 2999
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.925916972002858e-05,
+      "loss": 19.9292,
+      "step": 3000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.925839893156246e-05,
+      "loss": 18.1106,
+      "step": 3001
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.925762774532161e-05,
+      "loss": 15.0008,
+      "step": 3002
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.925685616131226e-05,
+      "loss": 17.3911,
+      "step": 3003
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.925608417954064e-05,
+      "loss": 17.4134,
+      "step": 3004
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.925531180001298e-05,
+      "loss": 17.078,
+      "step": 3005
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.925453902273553e-05,
+      "loss": 18.1953,
+      "step": 3006
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92537658477145e-05,
+      "loss": 18.0625,
+      "step": 3007
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.925299227495618e-05,
+      "loss": 15.1533,
+      "step": 3008
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.925221830446678e-05,
+      "loss": 16.7308,
+      "step": 3009
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.925144393625255e-05,
+      "loss": 17.229,
+      "step": 3010
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.925066917031977e-05,
+      "loss": 16.4237,
+      "step": 3011
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.924989400667468e-05,
+      "loss": 17.545,
+      "step": 3012
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.924911844532352e-05,
+      "loss": 16.0905,
+      "step": 3013
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92483424862726e-05,
+      "loss": 15.892,
+      "step": 3014
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.924756612952814e-05,
+      "loss": 15.7591,
+      "step": 3015
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.924678937509643e-05,
+      "loss": 17.447,
+      "step": 3016
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.924601222298375e-05,
+      "loss": 16.8782,
+      "step": 3017
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.924523467319636e-05,
+      "loss": 17.7227,
+      "step": 3018
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.924445672574055e-05,
+      "loss": 17.8851,
+      "step": 3019
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.924367838062259e-05,
+      "loss": 17.5928,
+      "step": 3020
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.924289963784879e-05,
+      "loss": 18.9641,
+      "step": 3021
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92421204974254e-05,
+      "loss": 15.4963,
+      "step": 3022
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.924134095935875e-05,
+      "loss": 15.1384,
+      "step": 3023
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.924056102365511e-05,
+      "loss": 16.9648,
+      "step": 3024
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.923978069032078e-05,
+      "loss": 15.901,
+      "step": 3025
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.923899995936209e-05,
+      "loss": 17.4201,
+      "step": 3026
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.923821883078531e-05,
+      "loss": 15.5283,
+      "step": 3027
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.923743730459676e-05,
+      "loss": 16.1718,
+      "step": 3028
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.923665538080275e-05,
+      "loss": 16.6194,
+      "step": 3029
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92358730594096e-05,
+      "loss": 17.3972,
+      "step": 3030
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.923509034042363e-05,
+      "loss": 15.8251,
+      "step": 3031
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.923430722385113e-05,
+      "loss": 17.7007,
+      "step": 3032
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.923352370969847e-05,
+      "loss": 16.6272,
+      "step": 3033
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.923273979797193e-05,
+      "loss": 17.0897,
+      "step": 3034
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92319554886779e-05,
+      "loss": 16.2314,
+      "step": 3035
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.923117078182264e-05,
+      "loss": 17.838,
+      "step": 3036
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.923038567741256e-05,
+      "loss": 16.6008,
+      "step": 3037
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.922960017545395e-05,
+      "loss": 17.4723,
+      "step": 3038
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.922881427595316e-05,
+      "loss": 17.6285,
+      "step": 3039
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.922802797891657e-05,
+      "loss": 18.6369,
+      "step": 3040
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.922724128435048e-05,
+      "loss": 15.8208,
+      "step": 3041
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.922645419226128e-05,
+      "loss": 16.8858,
+      "step": 3042
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92256667026553e-05,
+      "loss": 17.5703,
+      "step": 3043
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.922487881553892e-05,
+      "loss": 16.8349,
+      "step": 3044
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92240905309185e-05,
+      "loss": 17.4006,
+      "step": 3045
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92233018488004e-05,
+      "loss": 16.4113,
+      "step": 3046
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.922251276919098e-05,
+      "loss": 18.5284,
+      "step": 3047
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.922172329209663e-05,
+      "loss": 16.0372,
+      "step": 3048
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92209334175237e-05,
+      "loss": 16.0073,
+      "step": 3049
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92201431454786e-05,
+      "loss": 16.2993,
+      "step": 3050
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92193524759677e-05,
+      "loss": 18.9817,
+      "step": 3051
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.921856140899737e-05,
+      "loss": 18.1437,
+      "step": 3052
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.921776994457401e-05,
+      "loss": 17.3108,
+      "step": 3053
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.921697808270403e-05,
+      "loss": 17.5738,
+      "step": 3054
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.921618582339378e-05,
+      "loss": 17.1117,
+      "step": 3055
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.92153931666497e-05,
+      "loss": 18.9437,
+      "step": 3056
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.921460011247816e-05,
+      "loss": 17.238,
+      "step": 3057
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.921380666088558e-05,
+      "loss": 18.4444,
+      "step": 3058
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.921301281187838e-05,
+      "loss": 17.5258,
+      "step": 3059
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.921221856546293e-05,
+      "loss": 17.8073,
+      "step": 3060
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.921142392164568e-05,
+      "loss": 16.8383,
+      "step": 3061
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.921062888043305e-05,
+      "loss": 15.525,
+      "step": 3062
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.920983344183144e-05,
+      "loss": 16.0113,
+      "step": 3063
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.920903760584727e-05,
+      "loss": 19.3351,
+      "step": 3064
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.920824137248698e-05,
+      "loss": 16.499,
+      "step": 3065
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.9207444741757e-05,
+      "loss": 18.1361,
+      "step": 3066
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.920664771366378e-05,
+      "loss": 17.6427,
+      "step": 3067
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.92058502882137e-05,
+      "loss": 15.8496,
+      "step": 3068
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.920505246541325e-05,
+      "loss": 18.0066,
+      "step": 3069
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.920425424526886e-05,
+      "loss": 17.2764,
+      "step": 3070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.920345562778698e-05,
+      "loss": 15.7188,
+      "step": 3071
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.920265661297404e-05,
+      "loss": 18.378,
+      "step": 3072
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.920185720083652e-05,
+      "loss": 19.1302,
+      "step": 3073
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.920105739138083e-05,
+      "loss": 17.8756,
+      "step": 3074
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.920025718461349e-05,
+      "loss": 15.7602,
+      "step": 3075
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.919945658054092e-05,
+      "loss": 16.3815,
+      "step": 3076
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.91986555791696e-05,
+      "loss": 17.6367,
+      "step": 3077
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.919785418050598e-05,
+      "loss": 18.2027,
+      "step": 3078
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.919705238455656e-05,
+      "loss": 16.9368,
+      "step": 3079
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.919625019132778e-05,
+      "loss": 15.7155,
+      "step": 3080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.919544760082617e-05,
+      "loss": 16.9919,
+      "step": 3081
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.919464461305816e-05,
+      "loss": 18.4703,
+      "step": 3082
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.919384122803027e-05,
+      "loss": 16.86,
+      "step": 3083
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.919303744574895e-05,
+      "loss": 15.4156,
+      "step": 3084
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.919223326622072e-05,
+      "loss": 16.2297,
+      "step": 3085
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.919142868945207e-05,
+      "loss": 18.247,
+      "step": 3086
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.919062371544949e-05,
+      "loss": 17.0587,
+      "step": 3087
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.918981834421948e-05,
+      "loss": 16.4021,
+      "step": 3088
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.918901257576855e-05,
+      "loss": 18.1782,
+      "step": 3089
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.918820641010322e-05,
+      "loss": 15.5813,
+      "step": 3090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.918739984722995e-05,
+      "loss": 15.8449,
+      "step": 3091
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.918659288715533e-05,
+      "loss": 15.3688,
+      "step": 3092
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.91857855298858e-05,
+      "loss": 16.8158,
+      "step": 3093
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.918497777542793e-05,
+      "loss": 15.9055,
+      "step": 3094
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.918416962378823e-05,
+      "loss": 17.4148,
+      "step": 3095
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.91833610749732e-05,
+      "loss": 16.9075,
+      "step": 3096
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.91825521289894e-05,
+      "loss": 17.4517,
+      "step": 3097
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.918174278584336e-05,
+      "loss": 19.1156,
+      "step": 3098
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.918093304554161e-05,
+      "loss": 16.3794,
+      "step": 3099
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.918012290809068e-05,
+      "loss": 16.5398,
+      "step": 3100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.917931237349714e-05,
+      "loss": 15.5426,
+      "step": 3101
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.917850144176751e-05,
+      "loss": 18.7813,
+      "step": 3102
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.917769011290832e-05,
+      "loss": 15.5718,
+      "step": 3103
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.917687838692617e-05,
+      "loss": 16.5479,
+      "step": 3104
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.917606626382758e-05,
+      "loss": 17.9458,
+      "step": 3105
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.917525374361912e-05,
+      "loss": 16.2759,
+      "step": 3106
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.917444082630735e-05,
+      "loss": 16.5193,
+      "step": 3107
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.917362751189884e-05,
+      "loss": 18.9878,
+      "step": 3108
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.917281380040016e-05,
+      "loss": 17.9694,
+      "step": 3109
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.917199969181786e-05,
+      "loss": 16.1676,
+      "step": 3110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.917118518615854e-05,
+      "loss": 17.1868,
+      "step": 3111
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.917037028342874e-05,
+      "loss": 18.088,
+      "step": 3112
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.91695549836351e-05,
+      "loss": 17.0101,
+      "step": 3113
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.916873928678414e-05,
+      "loss": 17.9022,
+      "step": 3114
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.916792319288249e-05,
+      "loss": 16.0895,
+      "step": 3115
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.916710670193672e-05,
+      "loss": 17.187,
+      "step": 3116
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.916628981395344e-05,
+      "loss": 17.7502,
+      "step": 3117
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.916547252893923e-05,
+      "loss": 18.2089,
+      "step": 3118
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.916465484690069e-05,
+      "loss": 18.1867,
+      "step": 3119
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.916383676784444e-05,
+      "loss": 16.7521,
+      "step": 3120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.916301829177708e-05,
+      "loss": 18.1297,
+      "step": 3121
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.916219941870519e-05,
+      "loss": 18.8439,
+      "step": 3122
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.916138014863542e-05,
+      "loss": 16.3048,
+      "step": 3123
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.91605604815744e-05,
+      "loss": 20.0234,
+      "step": 3124
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.915974041752868e-05,
+      "loss": 20.6958,
+      "step": 3125
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.915891995650495e-05,
+      "loss": 17.2191,
+      "step": 3126
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.915809909850982e-05,
+      "loss": 16.5195,
+      "step": 3127
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.915727784354988e-05,
+      "loss": 17.0385,
+      "step": 3128
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.91564561916318e-05,
+      "loss": 16.818,
+      "step": 3129
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.915563414276221e-05,
+      "loss": 18.1699,
+      "step": 3130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.915481169694775e-05,
+      "loss": 17.0616,
+      "step": 3131
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.915398885419505e-05,
+      "loss": 16.039,
+      "step": 3132
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.915316561451076e-05,
+      "loss": 16.7837,
+      "step": 3133
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.915234197790152e-05,
+      "loss": 17.2317,
+      "step": 3134
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.9151517944374e-05,
+      "loss": 16.2438,
+      "step": 3135
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.915069351393484e-05,
+      "loss": 17.8101,
+      "step": 3136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.914986868659069e-05,
+      "loss": 17.8581,
+      "step": 3137
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.914904346234823e-05,
+      "loss": 16.5082,
+      "step": 3138
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.914821784121412e-05,
+      "loss": 17.2371,
+      "step": 3139
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.914739182319502e-05,
+      "loss": 18.0266,
+      "step": 3140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.91465654082976e-05,
+      "loss": 14.8868,
+      "step": 3141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.914573859652854e-05,
+      "loss": 17.2489,
+      "step": 3142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.91449113878945e-05,
+      "loss": 17.4182,
+      "step": 3143
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.914408378240218e-05,
+      "loss": 15.2851,
+      "step": 3144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.914325578005828e-05,
+      "loss": 15.6924,
+      "step": 3145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.914242738086943e-05,
+      "loss": 17.3474,
+      "step": 3146
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.914159858484237e-05,
+      "loss": 15.6649,
+      "step": 3147
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.914076939198376e-05,
+      "loss": 18.1771,
+      "step": 3148
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.913993980230031e-05,
+      "loss": 15.7787,
+      "step": 3149
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.913910981579875e-05,
+      "loss": 17.3365,
+      "step": 3150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.913827943248572e-05,
+      "loss": 18.2903,
+      "step": 3151
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.913744865236798e-05,
+      "loss": 16.1947,
+      "step": 3152
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.913661747545219e-05,
+      "loss": 16.7797,
+      "step": 3153
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.913578590174511e-05,
+      "loss": 18.1142,
+      "step": 3154
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.913495393125342e-05,
+      "loss": 18.0205,
+      "step": 3155
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.913412156398387e-05,
+      "loss": 17.694,
+      "step": 3156
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.913328879994315e-05,
+      "loss": 18.2035,
+      "step": 3157
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.9132455639138e-05,
+      "loss": 17.4636,
+      "step": 3158
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.913162208157516e-05,
+      "loss": 16.7901,
+      "step": 3159
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.913078812726134e-05,
+      "loss": 16.3952,
+      "step": 3160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.912995377620327e-05,
+      "loss": 17.9099,
+      "step": 3161
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.912911902840771e-05,
+      "loss": 17.1878,
+      "step": 3162
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.912828388388139e-05,
+      "loss": 16.8611,
+      "step": 3163
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.912744834263105e-05,
+      "loss": 17.6396,
+      "step": 3164
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.912661240466344e-05,
+      "loss": 17.3945,
+      "step": 3165
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.912577606998532e-05,
+      "loss": 17.2832,
+      "step": 3166
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.912493933860342e-05,
+      "loss": 18.3691,
+      "step": 3167
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.912410221052454e-05,
+      "loss": 17.3636,
+      "step": 3168
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.912326468575539e-05,
+      "loss": 15.4484,
+      "step": 3169
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.912242676430276e-05,
+      "loss": 15.6406,
+      "step": 3170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.912158844617344e-05,
+      "loss": 16.4473,
+      "step": 3171
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.912074973137412e-05,
+      "loss": 16.6386,
+      "step": 3172
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.911991061991165e-05,
+      "loss": 16.8857,
+      "step": 3173
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.911907111179279e-05,
+      "loss": 17.5187,
+      "step": 3174
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.911823120702432e-05,
+      "loss": 16.1349,
+      "step": 3175
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.911739090561298e-05,
+      "loss": 16.0613,
+      "step": 3176
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.91165502075656e-05,
+      "loss": 15.3712,
+      "step": 3177
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.911570911288896e-05,
+      "loss": 15.8968,
+      "step": 3178
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.911486762158986e-05,
+      "loss": 16.5647,
+      "step": 3179
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.911402573367508e-05,
+      "loss": 20.6421,
+      "step": 3180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.911318344915142e-05,
+      "loss": 18.3381,
+      "step": 3181
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.911234076802568e-05,
+      "loss": 17.9731,
+      "step": 3182
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.911149769030468e-05,
+      "loss": 16.0751,
+      "step": 3183
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.91106542159952e-05,
+      "loss": 18.7467,
+      "step": 3184
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.910981034510406e-05,
+      "loss": 15.8692,
+      "step": 3185
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.910896607763811e-05,
+      "loss": 18.9226,
+      "step": 3186
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.910812141360413e-05,
+      "loss": 17.0074,
+      "step": 3187
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.910727635300896e-05,
+      "loss": 16.8518,
+      "step": 3188
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.910643089585941e-05,
+      "loss": 15.9823,
+      "step": 3189
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.910558504216232e-05,
+      "loss": 16.3823,
+      "step": 3190
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.91047387919245e-05,
+      "loss": 18.2592,
+      "step": 3191
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.910389214515281e-05,
+      "loss": 16.2354,
+      "step": 3192
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.910304510185406e-05,
+      "loss": 17.648,
+      "step": 3193
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.910219766203512e-05,
+      "loss": 18.3474,
+      "step": 3194
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.91013498257028e-05,
+      "loss": 15.7829,
+      "step": 3195
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.910050159286399e-05,
+      "loss": 19.0991,
+      "step": 3196
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.909965296352549e-05,
+      "loss": 18.2481,
+      "step": 3197
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.90988039376942e-05,
+      "loss": 17.6517,
+      "step": 3198
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.909795451537694e-05,
+      "loss": 17.2777,
+      "step": 3199
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.909710469658056e-05,
+      "loss": 16.9311,
+      "step": 3200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.909625448131197e-05,
+      "loss": 18.9344,
+      "step": 3201
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.909540386957801e-05,
+      "loss": 16.5416,
+      "step": 3202
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.909455286138553e-05,
+      "loss": 15.9319,
+      "step": 3203
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.909370145674146e-05,
+      "loss": 15.6574,
+      "step": 3204
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.90928496556526e-05,
+      "loss": 17.6477,
+      "step": 3205
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.909199745812588e-05,
+      "loss": 16.5483,
+      "step": 3206
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.909114486416818e-05,
+      "loss": 17.1702,
+      "step": 3207
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.909029187378634e-05,
+      "loss": 17.9143,
+      "step": 3208
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.90894384869873e-05,
+      "loss": 18.1248,
+      "step": 3209
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.908858470377793e-05,
+      "loss": 19.1134,
+      "step": 3210
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.908773052416514e-05,
+      "loss": 17.1109,
+      "step": 3211
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.908687594815579e-05,
+      "loss": 16.6024,
+      "step": 3212
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.908602097575681e-05,
+      "loss": 17.4933,
+      "step": 3213
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.90851656069751e-05,
+      "loss": 17.5102,
+      "step": 3214
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.908430984181758e-05,
+      "loss": 17.4139,
+      "step": 3215
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.908345368029113e-05,
+      "loss": 18.1451,
+      "step": 3216
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.908259712240268e-05,
+      "loss": 17.2909,
+      "step": 3217
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.908174016815917e-05,
+      "loss": 17.3962,
+      "step": 3218
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.908088281756747e-05,
+      "loss": 16.6152,
+      "step": 3219
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.908002507063454e-05,
+      "loss": 18.6681,
+      "step": 3220
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.907916692736729e-05,
+      "loss": 18.1606,
+      "step": 3221
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.907830838777268e-05,
+      "loss": 16.9015,
+      "step": 3222
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.90774494518576e-05,
+      "loss": 18.3023,
+      "step": 3223
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.907659011962902e-05,
+      "loss": 16.4976,
+      "step": 3224
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.907573039109386e-05,
+      "loss": 16.2974,
+      "step": 3225
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.907487026625906e-05,
+      "loss": 17.833,
+      "step": 3226
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.907400974513158e-05,
+      "loss": 17.4841,
+      "step": 3227
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.907314882771836e-05,
+      "loss": 17.9948,
+      "step": 3228
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.907228751402636e-05,
+      "loss": 15.6763,
+      "step": 3229
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.907142580406253e-05,
+      "loss": 17.4091,
+      "step": 3230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.907056369783381e-05,
+      "loss": 16.6397,
+      "step": 3231
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.90697011953472e-05,
+      "loss": 16.8311,
+      "step": 3232
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.906883829660964e-05,
+      "loss": 18.0368,
+      "step": 3233
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.90679750016281e-05,
+      "loss": 16.9137,
+      "step": 3234
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.906711131040956e-05,
+      "loss": 16.3076,
+      "step": 3235
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.906624722296098e-05,
+      "loss": 16.0537,
+      "step": 3236
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.906538273928936e-05,
+      "loss": 16.6148,
+      "step": 3237
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.906451785940167e-05,
+      "loss": 19.3438,
+      "step": 3238
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.906365258330488e-05,
+      "loss": 18.4103,
+      "step": 3239
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.9062786911006e-05,
+      "loss": 16.8984,
+      "step": 3240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.9061920842512e-05,
+      "loss": 16.0737,
+      "step": 3241
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.90610543778299e-05,
+      "loss": 16.6195,
+      "step": 3242
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.906018751696667e-05,
+      "loss": 17.3997,
+      "step": 3243
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.905932025992932e-05,
+      "loss": 17.6018,
+      "step": 3244
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.905845260672484e-05,
+      "loss": 16.1451,
+      "step": 3245
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.905758455736028e-05,
+      "loss": 16.5962,
+      "step": 3246
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.905671611184259e-05,
+      "loss": 16.6218,
+      "step": 3247
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.905584727017884e-05,
+      "loss": 16.5929,
+      "step": 3248
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.905497803237601e-05,
+      "loss": 17.2013,
+      "step": 3249
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.905410839844113e-05,
+      "loss": 15.6405,
+      "step": 3250
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.905323836838121e-05,
+      "loss": 17.6291,
+      "step": 3251
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.90523679422033e-05,
+      "loss": 16.7419,
+      "step": 3252
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.905149711991442e-05,
+      "loss": 16.406,
+      "step": 3253
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.905062590152159e-05,
+      "loss": 17.9469,
+      "step": 3254
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.904975428703185e-05,
+      "loss": 18.3599,
+      "step": 3255
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.904888227645225e-05,
+      "loss": 17.7067,
+      "step": 3256
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.904800986978982e-05,
+      "loss": 15.2182,
+      "step": 3257
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.90471370670516e-05,
+      "loss": 16.1628,
+      "step": 3258
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.904626386824466e-05,
+      "loss": 16.3872,
+      "step": 3259
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.904539027337603e-05,
+      "loss": 17.4971,
+      "step": 3260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.904451628245277e-05,
+      "loss": 17.4777,
+      "step": 3261
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.904364189548195e-05,
+      "loss": 16.8576,
+      "step": 3262
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.904276711247062e-05,
+      "loss": 17.8923,
+      "step": 3263
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.904189193342583e-05,
+      "loss": 17.1093,
+      "step": 3264
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.904101635835468e-05,
+      "loss": 18.3859,
+      "step": 3265
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.904014038726422e-05,
+      "loss": 17.4339,
+      "step": 3266
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.903926402016153e-05,
+      "loss": 16.934,
+      "step": 3267
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.903838725705368e-05,
+      "loss": 18.1173,
+      "step": 3268
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.903751009794775e-05,
+      "loss": 16.6015,
+      "step": 3269
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.903663254285083e-05,
+      "loss": 16.5133,
+      "step": 3270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.903575459176999e-05,
+      "loss": 17.9127,
+      "step": 3271
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.903487624471236e-05,
+      "loss": 16.1771,
+      "step": 3272
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.903399750168498e-05,
+      "loss": 17.0609,
+      "step": 3273
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.903311836269498e-05,
+      "loss": 17.7379,
+      "step": 3274
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.903223882774946e-05,
+      "loss": 17.6527,
+      "step": 3275
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.903135889685551e-05,
+      "loss": 16.7655,
+      "step": 3276
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.903047857002025e-05,
+      "loss": 18.305,
+      "step": 3277
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.902959784725077e-05,
+      "loss": 17.2603,
+      "step": 3278
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.902871672855418e-05,
+      "loss": 17.1432,
+      "step": 3279
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.902783521393762e-05,
+      "loss": 18.5888,
+      "step": 3280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.902695330340819e-05,
+      "loss": 16.3554,
+      "step": 3281
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.902607099697301e-05,
+      "loss": 16.4784,
+      "step": 3282
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.902518829463921e-05,
+      "loss": 18.6542,
+      "step": 3283
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.902430519641392e-05,
+      "loss": 17.9408,
+      "step": 3284
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.902342170230428e-05,
+      "loss": 16.1277,
+      "step": 3285
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.90225378123174e-05,
+      "loss": 16.7347,
+      "step": 3286
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.902165352646045e-05,
+      "loss": 17.8893,
+      "step": 3287
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.902076884474054e-05,
+      "loss": 15.7946,
+      "step": 3288
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.901988376716483e-05,
+      "loss": 18.8117,
+      "step": 3289
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.901899829374047e-05,
+      "loss": 17.2136,
+      "step": 3290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.90181124244746e-05,
+      "loss": 16.458,
+      "step": 3291
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.901722615937438e-05,
+      "loss": 15.5593,
+      "step": 3292
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.901633949844698e-05,
+      "loss": 17.426,
+      "step": 3293
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.901545244169952e-05,
+      "loss": 18.8576,
+      "step": 3294
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.90145649891392e-05,
+      "loss": 16.7947,
+      "step": 3295
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.901367714077316e-05,
+      "loss": 19.0544,
+      "step": 3296
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.901278889660861e-05,
+      "loss": 18.1441,
+      "step": 3297
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.901190025665268e-05,
+      "loss": 15.3674,
+      "step": 3298
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.901101122091256e-05,
+      "loss": 16.822,
+      "step": 3299
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.901012178939545e-05,
+      "loss": 19.1104,
+      "step": 3300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.900923196210851e-05,
+      "loss": 17.4802,
+      "step": 3301
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.900834173905892e-05,
+      "loss": 18.2838,
+      "step": 3302
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.900745112025389e-05,
+      "loss": 15.2338,
+      "step": 3303
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.90065601057006e-05,
+      "loss": 16.4653,
+      "step": 3304
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.900566869540624e-05,
+      "loss": 19.6398,
+      "step": 3305
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.900477688937801e-05,
+      "loss": 16.5882,
+      "step": 3306
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.900388468762314e-05,
+      "loss": 18.1242,
+      "step": 3307
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.90029920901488e-05,
+      "loss": 14.7782,
+      "step": 3308
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.90020990969622e-05,
+      "loss": 17.406,
+      "step": 3309
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.900120570807057e-05,
+      "loss": 17.3419,
+      "step": 3310
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.90003119234811e-05,
+      "loss": 17.4036,
+      "step": 3311
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.899941774320103e-05,
+      "loss": 15.2832,
+      "step": 3312
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.899852316723759e-05,
+      "loss": 19.6839,
+      "step": 3313
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.899762819559797e-05,
+      "loss": 18.4044,
+      "step": 3314
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.899673282828941e-05,
+      "loss": 16.103,
+      "step": 3315
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.899583706531916e-05,
+      "loss": 16.437,
+      "step": 3316
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.899494090669442e-05,
+      "loss": 17.8387,
+      "step": 3317
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.899404435242246e-05,
+      "loss": 18.6219,
+      "step": 3318
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.899314740251049e-05,
+      "loss": 17.1904,
+      "step": 3319
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.899225005696578e-05,
+      "loss": 15.9665,
+      "step": 3320
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.899135231579556e-05,
+      "loss": 16.8066,
+      "step": 3321
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.899045417900709e-05,
+      "loss": 14.959,
+      "step": 3322
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89895556466076e-05,
+      "loss": 15.5785,
+      "step": 3323
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.898865671860438e-05,
+      "loss": 16.9937,
+      "step": 3324
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.898775739500466e-05,
+      "loss": 16.6713,
+      "step": 3325
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.898685767581572e-05,
+      "loss": 17.1082,
+      "step": 3326
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.898595756104481e-05,
+      "loss": 19.1967,
+      "step": 3327
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.898505705069921e-05,
+      "loss": 17.1586,
+      "step": 3328
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89841561447862e-05,
+      "loss": 16.475,
+      "step": 3329
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.898325484331303e-05,
+      "loss": 14.4865,
+      "step": 3330
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.898235314628702e-05,
+      "loss": 17.2775,
+      "step": 3331
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89814510537154e-05,
+      "loss": 16.4883,
+      "step": 3332
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.898054856560547e-05,
+      "loss": 18.7888,
+      "step": 3333
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.897964568196454e-05,
+      "loss": 15.7812,
+      "step": 3334
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.897874240279988e-05,
+      "loss": 16.8953,
+      "step": 3335
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.897783872811882e-05,
+      "loss": 15.6131,
+      "step": 3336
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.897693465792859e-05,
+      "loss": 17.4322,
+      "step": 3337
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.897603019223655e-05,
+      "loss": 15.3012,
+      "step": 3338
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.897512533104998e-05,
+      "loss": 17.3971,
+      "step": 3339
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89742200743762e-05,
+      "loss": 19.08,
+      "step": 3340
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.897331442222251e-05,
+      "loss": 18.0761,
+      "step": 3341
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89724083745962e-05,
+      "loss": 16.0403,
+      "step": 3342
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.897150193150464e-05,
+      "loss": 16.1548,
+      "step": 3343
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89705950929551e-05,
+      "loss": 15.7411,
+      "step": 3344
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.896968785895493e-05,
+      "loss": 17.8605,
+      "step": 3345
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.896878022951146e-05,
+      "loss": 17.4163,
+      "step": 3346
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.8967872204632e-05,
+      "loss": 16.7559,
+      "step": 3347
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89669637843239e-05,
+      "loss": 16.2692,
+      "step": 3348
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.896605496859448e-05,
+      "loss": 15.1885,
+      "step": 3349
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.896514575745108e-05,
+      "loss": 17.0839,
+      "step": 3350
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.896423615090106e-05,
+      "loss": 17.2049,
+      "step": 3351
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.896332614895174e-05,
+      "loss": 16.6169,
+      "step": 3352
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89624157516105e-05,
+      "loss": 18.4935,
+      "step": 3353
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.896150495888466e-05,
+      "loss": 18.2854,
+      "step": 3354
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89605937707816e-05,
+      "loss": 15.9381,
+      "step": 3355
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.895968218730866e-05,
+      "loss": 15.4803,
+      "step": 3356
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.895877020847322e-05,
+      "loss": 19.0612,
+      "step": 3357
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.895785783428262e-05,
+      "loss": 18.7751,
+      "step": 3358
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.895694506474423e-05,
+      "loss": 15.6056,
+      "step": 3359
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.895603189986545e-05,
+      "loss": 17.0386,
+      "step": 3360
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.895511833965363e-05,
+      "loss": 17.3349,
+      "step": 3361
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.895420438411616e-05,
+      "loss": 17.3795,
+      "step": 3362
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.895329003326039e-05,
+      "loss": 15.9522,
+      "step": 3363
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.895237528709374e-05,
+      "loss": 17.4681,
+      "step": 3364
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.895146014562358e-05,
+      "loss": 16.9762,
+      "step": 3365
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.895054460885731e-05,
+      "loss": 16.216,
+      "step": 3366
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.894962867680233e-05,
+      "loss": 15.7488,
+      "step": 3367
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.8948712349466e-05,
+      "loss": 17.0581,
+      "step": 3368
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.894779562685574e-05,
+      "loss": 15.1006,
+      "step": 3369
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.894687850897897e-05,
+      "loss": 19.0206,
+      "step": 3370
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.894596099584308e-05,
+      "loss": 16.7797,
+      "step": 3371
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.894504308745547e-05,
+      "loss": 16.9526,
+      "step": 3372
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.894412478382358e-05,
+      "loss": 17.2411,
+      "step": 3373
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89432060849548e-05,
+      "loss": 18.6217,
+      "step": 3374
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.894228699085655e-05,
+      "loss": 18.122,
+      "step": 3375
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.894136750153627e-05,
+      "loss": 14.4457,
+      "step": 3376
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.894044761700136e-05,
+      "loss": 16.8349,
+      "step": 3377
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.893952733725929e-05,
+      "loss": 18.535,
+      "step": 3378
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.893860666231743e-05,
+      "loss": 16.1541,
+      "step": 3379
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.893768559218328e-05,
+      "loss": 15.8375,
+      "step": 3380
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.893676412686422e-05,
+      "loss": 15.7826,
+      "step": 3381
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.893584226636772e-05,
+      "loss": 17.6707,
+      "step": 3382
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.893492001070123e-05,
+      "loss": 17.2481,
+      "step": 3383
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89339973598722e-05,
+      "loss": 17.076,
+      "step": 3384
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.893307431388805e-05,
+      "loss": 16.207,
+      "step": 3385
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.893215087275627e-05,
+      "loss": 18.0612,
+      "step": 3386
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89312270364843e-05,
+      "loss": 15.819,
+      "step": 3387
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89303028050796e-05,
+      "loss": 15.7242,
+      "step": 3388
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.892937817854962e-05,
+      "loss": 17.1614,
+      "step": 3389
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.892845315690185e-05,
+      "loss": 15.1924,
+      "step": 3390
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.892752774014375e-05,
+      "loss": 18.6003,
+      "step": 3391
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89266019282828e-05,
+      "loss": 18.2226,
+      "step": 3392
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.892567572132647e-05,
+      "loss": 16.031,
+      "step": 3393
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.892474911928225e-05,
+      "loss": 17.031,
+      "step": 3394
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.892382212215758e-05,
+      "loss": 16.3277,
+      "step": 3395
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.892289472996002e-05,
+      "loss": 18.0392,
+      "step": 3396
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.892196694269697e-05,
+      "loss": 17.9005,
+      "step": 3397
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.8921038760376e-05,
+      "loss": 16.5154,
+      "step": 3398
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.892011018300456e-05,
+      "loss": 17.964,
+      "step": 3399
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.891918121059019e-05,
+      "loss": 16.0659,
+      "step": 3400
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.891825184314033e-05,
+      "loss": 18.4625,
+      "step": 3401
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.891732208066253e-05,
+      "loss": 16.8829,
+      "step": 3402
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.891639192316429e-05,
+      "loss": 18.3658,
+      "step": 3403
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.891546137065313e-05,
+      "loss": 16.9744,
+      "step": 3404
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.891453042313655e-05,
+      "loss": 17.7583,
+      "step": 3405
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.891359908062206e-05,
+      "loss": 17.9318,
+      "step": 3406
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.891266734311719e-05,
+      "loss": 17.017,
+      "step": 3407
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.891173521062947e-05,
+      "loss": 16.7491,
+      "step": 3408
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.891080268316642e-05,
+      "loss": 16.9615,
+      "step": 3409
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.890986976073557e-05,
+      "loss": 15.8265,
+      "step": 3410
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.890893644334447e-05,
+      "loss": 15.4159,
+      "step": 3411
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.890800273100063e-05,
+      "loss": 15.603,
+      "step": 3412
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.890706862371162e-05,
+      "loss": 17.9751,
+      "step": 3413
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.890613412148495e-05,
+      "loss": 17.1275,
+      "step": 3414
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.89051992243282e-05,
+      "loss": 17.6054,
+      "step": 3415
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.890426393224889e-05,
+      "loss": 15.0132,
+      "step": 3416
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.890332824525458e-05,
+      "loss": 18.8421,
+      "step": 3417
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.890239216335284e-05,
+      "loss": 15.5833,
+      "step": 3418
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.890145568655123e-05,
+      "loss": 16.6957,
+      "step": 3419
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.890051881485729e-05,
+      "loss": 19.789,
+      "step": 3420
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.88995815482786e-05,
+      "loss": 17.1435,
+      "step": 3421
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.889864388682273e-05,
+      "loss": 17.8045,
+      "step": 3422
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.889770583049724e-05,
+      "loss": 16.6187,
+      "step": 3423
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.889676737930972e-05,
+      "loss": 16.5528,
+      "step": 3424
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.889582853326776e-05,
+      "loss": 15.8604,
+      "step": 3425
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.889488929237889e-05,
+      "loss": 18.1533,
+      "step": 3426
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.889394965665075e-05,
+      "loss": 18.2953,
+      "step": 3427
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.88930096260909e-05,
+      "loss": 17.8639,
+      "step": 3428
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.889206920070693e-05,
+      "loss": 16.2267,
+      "step": 3429
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.889112838050646e-05,
+      "loss": 16.7135,
+      "step": 3430
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.889018716549705e-05,
+      "loss": 16.2494,
+      "step": 3431
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.888924555568632e-05,
+      "loss": 18.0792,
+      "step": 3432
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.88883035510819e-05,
+      "loss": 18.142,
+      "step": 3433
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.888736115169135e-05,
+      "loss": 17.4136,
+      "step": 3434
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.888641835752229e-05,
+      "loss": 15.5911,
+      "step": 3435
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.888547516858235e-05,
+      "loss": 16.5487,
+      "step": 3436
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.888453158487915e-05,
+      "loss": 15.3854,
+      "step": 3437
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.888358760642029e-05,
+      "loss": 15.7708,
+      "step": 3438
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.88826432332134e-05,
+      "loss": 17.054,
+      "step": 3439
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.888169846526611e-05,
+      "loss": 16.5515,
+      "step": 3440
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.888075330258606e-05,
+      "loss": 14.532,
+      "step": 3441
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.887980774518085e-05,
+      "loss": 17.5042,
+      "step": 3442
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.887886179305814e-05,
+      "loss": 17.4959,
+      "step": 3443
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.887791544622557e-05,
+      "loss": 15.8948,
+      "step": 3444
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.887696870469078e-05,
+      "loss": 18.1034,
+      "step": 3445
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.88760215684614e-05,
+      "loss": 17.5805,
+      "step": 3446
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.887507403754512e-05,
+      "loss": 15.5082,
+      "step": 3447
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.887412611194953e-05,
+      "loss": 15.9651,
+      "step": 3448
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.887317779168232e-05,
+      "loss": 18.9576,
+      "step": 3449
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.887222907675115e-05,
+      "loss": 16.4633,
+      "step": 3450
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.887127996716369e-05,
+      "loss": 17.2659,
+      "step": 3451
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.887033046292758e-05,
+      "loss": 17.1369,
+      "step": 3452
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.886938056405048e-05,
+      "loss": 17.0663,
+      "step": 3453
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.886843027054009e-05,
+      "loss": 16.3793,
+      "step": 3454
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.886747958240408e-05,
+      "loss": 16.9312,
+      "step": 3455
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.886652849965012e-05,
+      "loss": 16.6948,
+      "step": 3456
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.886557702228587e-05,
+      "loss": 16.8794,
+      "step": 3457
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.886462515031905e-05,
+      "loss": 17.2407,
+      "step": 3458
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.886367288375732e-05,
+      "loss": 17.9091,
+      "step": 3459
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.886272022260839e-05,
+      "loss": 15.3136,
+      "step": 3460
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.886176716687992e-05,
+      "loss": 15.6419,
+      "step": 3461
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.886081371657964e-05,
+      "loss": 17.1601,
+      "step": 3462
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.885985987171525e-05,
+      "loss": 16.0336,
+      "step": 3463
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.885890563229443e-05,
+      "loss": 18.2323,
+      "step": 3464
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.88579509983249e-05,
+      "loss": 16.8767,
+      "step": 3465
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.885699596981434e-05,
+      "loss": 18.4581,
+      "step": 3466
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.885604054677052e-05,
+      "loss": 18.1846,
+      "step": 3467
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.885508472920111e-05,
+      "loss": 18.7336,
+      "step": 3468
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.885412851711385e-05,
+      "loss": 16.3024,
+      "step": 3469
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.885317191051645e-05,
+      "loss": 18.2598,
+      "step": 3470
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.88522149094166e-05,
+      "loss": 16.2245,
+      "step": 3471
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.885125751382211e-05,
+      "loss": 16.8762,
+      "step": 3472
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.885029972374065e-05,
+      "loss": 16.876,
+      "step": 3473
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.884934153917997e-05,
+      "loss": 18.6226,
+      "step": 3474
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.88483829601478e-05,
+      "loss": 18.6562,
+      "step": 3475
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.884742398665191e-05,
+      "loss": 16.1374,
+      "step": 3476
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.884646461870002e-05,
+      "loss": 17.4159,
+      "step": 3477
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.884550485629987e-05,
+      "loss": 18.6801,
+      "step": 3478
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.884454469945922e-05,
+      "loss": 16.9883,
+      "step": 3479
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.884358414818583e-05,
+      "loss": 14.1074,
+      "step": 3480
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.884262320248745e-05,
+      "loss": 17.4112,
+      "step": 3481
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.884166186237183e-05,
+      "loss": 17.5779,
+      "step": 3482
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.884070012784677e-05,
+      "loss": 16.6394,
+      "step": 3483
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.883973799891998e-05,
+      "loss": 15.6091,
+      "step": 3484
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.883877547559928e-05,
+      "loss": 18.4687,
+      "step": 3485
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.883781255789242e-05,
+      "loss": 15.4216,
+      "step": 3486
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.883684924580715e-05,
+      "loss": 17.3968,
+      "step": 3487
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.883588553935132e-05,
+      "loss": 18.9309,
+      "step": 3488
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.883492143853265e-05,
+      "loss": 17.221,
+      "step": 3489
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.883395694335894e-05,
+      "loss": 15.8384,
+      "step": 3490
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.883299205383798e-05,
+      "loss": 15.7901,
+      "step": 3491
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.883202676997758e-05,
+      "loss": 16.3522,
+      "step": 3492
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.883106109178551e-05,
+      "loss": 16.6243,
+      "step": 3493
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.883009501926958e-05,
+      "loss": 18.9645,
+      "step": 3494
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.882912855243759e-05,
+      "loss": 17.4965,
+      "step": 3495
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.882816169129734e-05,
+      "loss": 18.6523,
+      "step": 3496
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.882719443585664e-05,
+      "loss": 18.9775,
+      "step": 3497
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.882622678612332e-05,
+      "loss": 18.5718,
+      "step": 3498
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.882525874210515e-05,
+      "loss": 17.0529,
+      "step": 3499
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.882429030380999e-05,
+      "loss": 15.9213,
+      "step": 3500
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.882332147124565e-05,
+      "loss": 17.2964,
+      "step": 3501
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.882235224441994e-05,
+      "loss": 17.095,
+      "step": 3502
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.882138262334068e-05,
+      "loss": 16.5782,
+      "step": 3503
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.882041260801574e-05,
+      "loss": 17.8785,
+      "step": 3504
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.881944219845291e-05,
+      "loss": 19.2864,
+      "step": 3505
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.881847139466004e-05,
+      "loss": 17.2723,
+      "step": 3506
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.881750019664498e-05,
+      "loss": 17.7389,
+      "step": 3507
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.881652860441557e-05,
+      "loss": 17.1241,
+      "step": 3508
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.881555661797965e-05,
+      "loss": 15.7567,
+      "step": 3509
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.881458423734506e-05,
+      "loss": 17.0391,
+      "step": 3510
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.881361146251967e-05,
+      "loss": 14.9993,
+      "step": 3511
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.881263829351132e-05,
+      "loss": 17.3498,
+      "step": 3512
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.88116647303279e-05,
+      "loss": 17.1536,
+      "step": 3513
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.881069077297723e-05,
+      "loss": 16.4754,
+      "step": 3514
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.88097164214672e-05,
+      "loss": 17.3751,
+      "step": 3515
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.880874167580567e-05,
+      "loss": 17.4087,
+      "step": 3516
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.880776653600052e-05,
+      "loss": 14.3607,
+      "step": 3517
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.88067910020596e-05,
+      "loss": 16.3045,
+      "step": 3518
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.880581507399083e-05,
+      "loss": 18.3987,
+      "step": 3519
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.880483875180205e-05,
+      "loss": 19.1797,
+      "step": 3520
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.880386203550116e-05,
+      "loss": 18.1439,
+      "step": 3521
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.880288492509606e-05,
+      "loss": 18.1518,
+      "step": 3522
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.88019074205946e-05,
+      "loss": 16.436,
+      "step": 3523
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.880092952200473e-05,
+      "loss": 17.2078,
+      "step": 3524
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.879995122933431e-05,
+      "loss": 16.261,
+      "step": 3525
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.879897254259124e-05,
+      "loss": 18.9331,
+      "step": 3526
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.879799346178344e-05,
+      "loss": 17.5545,
+      "step": 3527
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.879701398691882e-05,
+      "loss": 17.7938,
+      "step": 3528
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.879603411800526e-05,
+      "loss": 15.5729,
+      "step": 3529
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.87950538550507e-05,
+      "loss": 15.3117,
+      "step": 3530
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.879407319806304e-05,
+      "loss": 17.0552,
+      "step": 3531
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.87930921470502e-05,
+      "loss": 15.1733,
+      "step": 3532
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.879211070202011e-05,
+      "loss": 16.7129,
+      "step": 3533
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.87911288629807e-05,
+      "loss": 18.8697,
+      "step": 3534
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.879014662993989e-05,
+      "loss": 16.6324,
+      "step": 3535
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.878916400290561e-05,
+      "loss": 17.0898,
+      "step": 3536
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.878818098188579e-05,
+      "loss": 16.49,
+      "step": 3537
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.878719756688839e-05,
+      "loss": 18.9322,
+      "step": 3538
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.878621375792133e-05,
+      "loss": 16.3045,
+      "step": 3539
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.878522955499257e-05,
+      "loss": 17.704,
+      "step": 3540
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.878424495811004e-05,
+      "loss": 18.8738,
+      "step": 3541
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.87832599672817e-05,
+      "loss": 18.2146,
+      "step": 3542
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.878227458251552e-05,
+      "loss": 18.4464,
+      "step": 3543
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.878128880381943e-05,
+      "loss": 16.3906,
+      "step": 3544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.878030263120139e-05,
+      "loss": 16.1037,
+      "step": 3545
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.877931606466939e-05,
+      "loss": 16.9896,
+      "step": 3546
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.87783291042314e-05,
+      "loss": 16.4124,
+      "step": 3547
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.877734174989534e-05,
+      "loss": 18.8875,
+      "step": 3548
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.877635400166924e-05,
+      "loss": 15.7952,
+      "step": 3549
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.877536585956105e-05,
+      "loss": 17.867,
+      "step": 3550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.877437732357874e-05,
+      "loss": 17.8281,
+      "step": 3551
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.877338839373032e-05,
+      "loss": 15.5119,
+      "step": 3552
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.877239907002374e-05,
+      "loss": 16.5019,
+      "step": 3553
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.877140935246702e-05,
+      "loss": 15.9778,
+      "step": 3554
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.877041924106814e-05,
+      "loss": 14.9391,
+      "step": 3555
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.87694287358351e-05,
+      "loss": 16.0898,
+      "step": 3556
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.87684378367759e-05,
+      "loss": 15.9492,
+      "step": 3557
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.876744654389854e-05,
+      "loss": 15.486,
+      "step": 3558
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.8766454857211e-05,
+      "loss": 16.7712,
+      "step": 3559
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.876546277672132e-05,
+      "loss": 15.8288,
+      "step": 3560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.876447030243752e-05,
+      "loss": 17.7132,
+      "step": 3561
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.876347743436758e-05,
+      "loss": 17.9627,
+      "step": 3562
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.876248417251953e-05,
+      "loss": 16.6871,
+      "step": 3563
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.87614905169014e-05,
+      "loss": 17.9591,
+      "step": 3564
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.876049646752121e-05,
+      "loss": 17.0852,
+      "step": 3565
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.8759502024387e-05,
+      "loss": 20.0951,
+      "step": 3566
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.875850718750676e-05,
+      "loss": 17.7788,
+      "step": 3567
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.875751195688857e-05,
+      "loss": 17.5649,
+      "step": 3568
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.875651633254045e-05,
+      "loss": 18.2554,
+      "step": 3569
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.875552031447042e-05,
+      "loss": 18.022,
+      "step": 3570
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.875452390268653e-05,
+      "loss": 16.1738,
+      "step": 3571
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.875352709719687e-05,
+      "loss": 17.7437,
+      "step": 3572
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.875252989800944e-05,
+      "loss": 16.0444,
+      "step": 3573
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.87515323051323e-05,
+      "loss": 18.5224,
+      "step": 3574
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.875053431857353e-05,
+      "loss": 17.4132,
+      "step": 3575
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.874953593834115e-05,
+      "loss": 16.0717,
+      "step": 3576
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.874853716444327e-05,
+      "loss": 15.7342,
+      "step": 3577
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.874753799688792e-05,
+      "loss": 17.9452,
+      "step": 3578
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.874653843568317e-05,
+      "loss": 19.8802,
+      "step": 3579
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.874553848083711e-05,
+      "loss": 15.2644,
+      "step": 3580
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.874453813235782e-05,
+      "loss": 15.7958,
+      "step": 3581
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.874353739025335e-05,
+      "loss": 16.1789,
+      "step": 3582
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.87425362545318e-05,
+      "loss": 15.8977,
+      "step": 3583
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.874153472520125e-05,
+      "loss": 16.0488,
+      "step": 3584
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.874053280226977e-05,
+      "loss": 18.0395,
+      "step": 3585
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.873953048574548e-05,
+      "loss": 17.9963,
+      "step": 3586
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.873852777563646e-05,
+      "loss": 16.4591,
+      "step": 3587
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.873752467195081e-05,
+      "loss": 17.4907,
+      "step": 3588
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.873652117469665e-05,
+      "loss": 17.2334,
+      "step": 3589
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.873551728388203e-05,
+      "loss": 16.8078,
+      "step": 3590
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.873451299951511e-05,
+      "loss": 16.3836,
+      "step": 3591
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.873350832160397e-05,
+      "loss": 17.2571,
+      "step": 3592
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.873250325015673e-05,
+      "loss": 18.0275,
+      "step": 3593
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.873149778518151e-05,
+      "loss": 18.1427,
+      "step": 3594
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.873049192668644e-05,
+      "loss": 19.2344,
+      "step": 3595
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.872948567467963e-05,
+      "loss": 17.1281,
+      "step": 3596
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.872847902916919e-05,
+      "loss": 17.5344,
+      "step": 3597
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.872747199016328e-05,
+      "loss": 18.1715,
+      "step": 3598
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.872646455767002e-05,
+      "loss": 17.8564,
+      "step": 3599
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.872545673169753e-05,
+      "loss": 16.5889,
+      "step": 3600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.872444851225397e-05,
+      "loss": 16.1668,
+      "step": 3601
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.872343989934747e-05,
+      "loss": 16.2017,
+      "step": 3602
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.872243089298619e-05,
+      "loss": 17.3832,
+      "step": 3603
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.872142149317824e-05,
+      "loss": 17.8236,
+      "step": 3604
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.872041169993183e-05,
+      "loss": 15.9304,
+      "step": 3605
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.871940151325505e-05,
+      "loss": 17.6185,
+      "step": 3606
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.87183909331561e-05,
+      "loss": 16.8047,
+      "step": 3607
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.871737995964315e-05,
+      "loss": 18.2601,
+      "step": 3608
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.87163685927243e-05,
+      "loss": 16.7872,
+      "step": 3609
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.87153568324078e-05,
+      "loss": 17.7504,
+      "step": 3610
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.871434467870175e-05,
+      "loss": 16.6648,
+      "step": 3611
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.871333213161438e-05,
+      "loss": 17.5165,
+      "step": 3612
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.871231919115383e-05,
+      "loss": 17.0488,
+      "step": 3613
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.87113058573283e-05,
+      "loss": 15.4209,
+      "step": 3614
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.871029213014596e-05,
+      "loss": 16.5325,
+      "step": 3615
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.870927800961498e-05,
+      "loss": 16.8985,
+      "step": 3616
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.870826349574358e-05,
+      "loss": 17.1933,
+      "step": 3617
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.870724858853995e-05,
+      "loss": 16.0744,
+      "step": 3618
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.870623328801227e-05,
+      "loss": 16.7581,
+      "step": 3619
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.870521759416876e-05,
+      "loss": 18.3207,
+      "step": 3620
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.870420150701758e-05,
+      "loss": 15.7931,
+      "step": 3621
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.870318502656698e-05,
+      "loss": 15.0214,
+      "step": 3622
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.870216815282516e-05,
+      "loss": 14.9711,
+      "step": 3623
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.87011508858003e-05,
+      "loss": 16.8689,
+      "step": 3624
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.870013322550065e-05,
+      "loss": 18.3466,
+      "step": 3625
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.869911517193442e-05,
+      "loss": 16.8615,
+      "step": 3626
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.869809672510981e-05,
+      "loss": 17.1435,
+      "step": 3627
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.869707788503508e-05,
+      "loss": 15.563,
+      "step": 3628
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.869605865171843e-05,
+      "loss": 16.6137,
+      "step": 3629
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.86950390251681e-05,
+      "loss": 16.1638,
+      "step": 3630
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.869401900539233e-05,
+      "loss": 15.5861,
+      "step": 3631
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.869299859239933e-05,
+      "loss": 18.5023,
+      "step": 3632
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.869197778619738e-05,
+      "loss": 16.7579,
+      "step": 3633
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.869095658679469e-05,
+      "loss": 16.4788,
+      "step": 3634
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.868993499419951e-05,
+      "loss": 18.1943,
+      "step": 3635
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.868891300842013e-05,
+      "loss": 17.5815,
+      "step": 3636
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.868789062946475e-05,
+      "loss": 14.8354,
+      "step": 3637
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.868686785734165e-05,
+      "loss": 16.443,
+      "step": 3638
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.868584469205908e-05,
+      "loss": 17.0394,
+      "step": 3639
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.868482113362532e-05,
+      "loss": 17.6533,
+      "step": 3640
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.868379718204862e-05,
+      "loss": 16.5687,
+      "step": 3641
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.868277283733726e-05,
+      "loss": 17.3304,
+      "step": 3642
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.86817480994995e-05,
+      "loss": 17.6644,
+      "step": 3643
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.868072296854362e-05,
+      "loss": 15.7571,
+      "step": 3644
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.86796974444779e-05,
+      "loss": 16.9193,
+      "step": 3645
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.867867152731061e-05,
+      "loss": 16.4905,
+      "step": 3646
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.867764521705005e-05,
+      "loss": 16.4177,
+      "step": 3647
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.867661851370452e-05,
+      "loss": 16.1657,
+      "step": 3648
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.867559141728227e-05,
+      "loss": 17.0507,
+      "step": 3649
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.867456392779161e-05,
+      "loss": 16.7907,
+      "step": 3650
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.867353604524088e-05,
+      "loss": 17.2956,
+      "step": 3651
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.867250776963833e-05,
+      "loss": 17.3325,
+      "step": 3652
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.867147910099225e-05,
+      "loss": 15.0,
+      "step": 3653
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.867045003931101e-05,
+      "loss": 16.6657,
+      "step": 3654
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.866942058460287e-05,
+      "loss": 17.7223,
+      "step": 3655
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.866839073687616e-05,
+      "loss": 16.9817,
+      "step": 3656
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.866736049613919e-05,
+      "loss": 18.3444,
+      "step": 3657
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.86663298624003e-05,
+      "loss": 16.6434,
+      "step": 3658
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.866529883566778e-05,
+      "loss": 16.9101,
+      "step": 3659
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.866426741594998e-05,
+      "loss": 17.602,
+      "step": 3660
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.866323560325522e-05,
+      "loss": 17.2569,
+      "step": 3661
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.866220339759183e-05,
+      "loss": 17.7771,
+      "step": 3662
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.866117079896815e-05,
+      "loss": 18.0306,
+      "step": 3663
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.866013780739253e-05,
+      "loss": 17.3769,
+      "step": 3664
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.865910442287328e-05,
+      "loss": 15.8439,
+      "step": 3665
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.865807064541877e-05,
+      "loss": 17.3006,
+      "step": 3666
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.865703647503735e-05,
+      "loss": 17.9055,
+      "step": 3667
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.865600191173736e-05,
+      "loss": 16.5242,
+      "step": 3668
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.865496695552716e-05,
+      "loss": 17.5049,
+      "step": 3669
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.865393160641509e-05,
+      "loss": 16.2996,
+      "step": 3670
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.865289586440954e-05,
+      "loss": 14.6613,
+      "step": 3671
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.865185972951886e-05,
+      "loss": 14.7621,
+      "step": 3672
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.865082320175141e-05,
+      "loss": 16.0265,
+      "step": 3673
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.864978628111556e-05,
+      "loss": 17.6669,
+      "step": 3674
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.86487489676197e-05,
+      "loss": 16.8526,
+      "step": 3675
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.864771126127222e-05,
+      "loss": 17.5426,
+      "step": 3676
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.864667316208145e-05,
+      "loss": 17.9783,
+      "step": 3677
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.86456346700558e-05,
+      "loss": 17.5698,
+      "step": 3678
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.864459578520367e-05,
+      "loss": 15.1003,
+      "step": 3679
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.864355650753341e-05,
+      "loss": 16.6074,
+      "step": 3680
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.864251683705346e-05,
+      "loss": 16.9588,
+      "step": 3681
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.864147677377217e-05,
+      "loss": 17.9379,
+      "step": 3682
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.8640436317698e-05,
+      "loss": 16.48,
+      "step": 3683
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.863939546883928e-05,
+      "loss": 17.1462,
+      "step": 3684
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.863835422720447e-05,
+      "loss": 18.3848,
+      "step": 3685
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.863731259280193e-05,
+      "loss": 16.8287,
+      "step": 3686
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.863627056564013e-05,
+      "loss": 17.419,
+      "step": 3687
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.863522814572743e-05,
+      "loss": 17.4349,
+      "step": 3688
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.86341853330723e-05,
+      "loss": 16.1197,
+      "step": 3689
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.86331421276831e-05,
+      "loss": 16.0853,
+      "step": 3690
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.86320985295683e-05,
+      "loss": 19.1074,
+      "step": 3691
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.863105453873632e-05,
+      "loss": 16.3135,
+      "step": 3692
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.863001015519557e-05,
+      "loss": 18.2238,
+      "step": 3693
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.86289653789545e-05,
+      "loss": 16.1796,
+      "step": 3694
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.862792021002156e-05,
+      "loss": 17.6858,
+      "step": 3695
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.862687464840516e-05,
+      "loss": 16.2513,
+      "step": 3696
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.862582869411376e-05,
+      "loss": 16.2108,
+      "step": 3697
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.862478234715582e-05,
+      "loss": 17.6196,
+      "step": 3698
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.862373560753976e-05,
+      "loss": 17.5964,
+      "step": 3699
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.862268847527404e-05,
+      "loss": 16.0819,
+      "step": 3700
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.862164095036714e-05,
+      "loss": 17.536,
+      "step": 3701
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.862059303282749e-05,
+      "loss": 16.4684,
+      "step": 3702
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.861954472266357e-05,
+      "loss": 15.8441,
+      "step": 3703
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.861849601988383e-05,
+      "loss": 17.6674,
+      "step": 3704
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.861744692449675e-05,
+      "loss": 15.5672,
+      "step": 3705
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.86163974365108e-05,
+      "loss": 16.0289,
+      "step": 3706
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.861534755593446e-05,
+      "loss": 18.3712,
+      "step": 3707
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.86142972827762e-05,
+      "loss": 17.8501,
+      "step": 3708
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.861324661704451e-05,
+      "loss": 17.7747,
+      "step": 3709
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.861219555874785e-05,
+      "loss": 15.8532,
+      "step": 3710
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.861114410789475e-05,
+      "loss": 16.8152,
+      "step": 3711
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.861009226449366e-05,
+      "loss": 16.1416,
+      "step": 3712
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.86090400285531e-05,
+      "loss": 16.2508,
+      "step": 3713
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.860798740008155e-05,
+      "loss": 18.4463,
+      "step": 3714
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.860693437908751e-05,
+      "loss": 15.3082,
+      "step": 3715
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.86058809655795e-05,
+      "loss": 17.1743,
+      "step": 3716
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.860482715956601e-05,
+      "loss": 17.648,
+      "step": 3717
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.860377296105556e-05,
+      "loss": 16.3911,
+      "step": 3718
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.860271837005668e-05,
+      "loss": 18.3289,
+      "step": 3719
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.860166338657783e-05,
+      "loss": 15.8422,
+      "step": 3720
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.86006080106276e-05,
+      "loss": 16.8158,
+      "step": 3721
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.859955224221446e-05,
+      "loss": 15.9834,
+      "step": 3722
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.859849608134697e-05,
+      "loss": 15.9837,
+      "step": 3723
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.859743952803362e-05,
+      "loss": 15.573,
+      "step": 3724
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.859638258228298e-05,
+      "loss": 18.271,
+      "step": 3725
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.859532524410357e-05,
+      "loss": 18.0839,
+      "step": 3726
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.859426751350391e-05,
+      "loss": 15.7761,
+      "step": 3727
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.859320939049257e-05,
+      "loss": 16.6637,
+      "step": 3728
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.859215087507809e-05,
+      "loss": 17.6841,
+      "step": 3729
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.859109196726901e-05,
+      "loss": 17.2923,
+      "step": 3730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.859003266707387e-05,
+      "loss": 16.7999,
+      "step": 3731
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.858897297450126e-05,
+      "loss": 17.2318,
+      "step": 3732
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.858791288955969e-05,
+      "loss": 16.1302,
+      "step": 3733
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.858685241225776e-05,
+      "loss": 15.391,
+      "step": 3734
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.858579154260401e-05,
+      "loss": 16.9242,
+      "step": 3735
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.858473028060703e-05,
+      "loss": 16.8155,
+      "step": 3736
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.858366862627536e-05,
+      "loss": 15.3904,
+      "step": 3737
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.85826065796176e-05,
+      "loss": 16.2979,
+      "step": 3738
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.85815441406423e-05,
+      "loss": 16.1231,
+      "step": 3739
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.858048130935806e-05,
+      "loss": 18.7076,
+      "step": 3740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.857941808577347e-05,
+      "loss": 17.7139,
+      "step": 3741
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.857835446989707e-05,
+      "loss": 18.1966,
+      "step": 3742
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.857729046173751e-05,
+      "loss": 15.6268,
+      "step": 3743
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.857622606130334e-05,
+      "loss": 16.0008,
+      "step": 3744
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.857516126860318e-05,
+      "loss": 18.1653,
+      "step": 3745
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.85740960836456e-05,
+      "loss": 15.1066,
+      "step": 3746
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.857303050643925e-05,
+      "loss": 16.6721,
+      "step": 3747
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.857196453699266e-05,
+      "loss": 16.7328,
+      "step": 3748
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.857089817531452e-05,
+      "loss": 14.6809,
+      "step": 3749
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.856983142141339e-05,
+      "loss": 17.531,
+      "step": 3750
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.856876427529788e-05,
+      "loss": 17.5451,
+      "step": 3751
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.856769673697666e-05,
+      "loss": 16.3976,
+      "step": 3752
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.856662880645828e-05,
+      "loss": 18.4619,
+      "step": 3753
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.856556048375143e-05,
+      "loss": 16.6566,
+      "step": 3754
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.856449176886468e-05,
+      "loss": 19.4542,
+      "step": 3755
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.856342266180672e-05,
+      "loss": 18.8633,
+      "step": 3756
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.856235316258611e-05,
+      "loss": 17.2365,
+      "step": 3757
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.856128327121155e-05,
+      "loss": 17.9536,
+      "step": 3758
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.856021298769164e-05,
+      "loss": 15.8271,
+      "step": 3759
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.855914231203506e-05,
+      "loss": 16.623,
+      "step": 3760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.855807124425042e-05,
+      "loss": 16.6817,
+      "step": 3761
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.855699978434639e-05,
+      "loss": 16.4387,
+      "step": 3762
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.855592793233162e-05,
+      "loss": 18.6336,
+      "step": 3763
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.855485568821475e-05,
+      "loss": 17.1173,
+      "step": 3764
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.855378305200444e-05,
+      "loss": 16.2349,
+      "step": 3765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.855271002370938e-05,
+      "loss": 17.6129,
+      "step": 3766
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.855163660333823e-05,
+      "loss": 17.2765,
+      "step": 3767
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.855056279089963e-05,
+      "loss": 16.1461,
+      "step": 3768
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.854948858640226e-05,
+      "loss": 16.3338,
+      "step": 3769
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.854841398985481e-05,
+      "loss": 18.6092,
+      "step": 3770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.854733900126595e-05,
+      "loss": 19.383,
+      "step": 3771
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.854626362064436e-05,
+      "loss": 16.634,
+      "step": 3772
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.854518784799871e-05,
+      "loss": 17.9125,
+      "step": 3773
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.854411168333772e-05,
+      "loss": 18.9578,
+      "step": 3774
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.854303512667004e-05,
+      "loss": 18.3007,
+      "step": 3775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.85419581780044e-05,
+      "loss": 17.1543,
+      "step": 3776
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.854088083734949e-05,
+      "loss": 17.4088,
+      "step": 3777
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.853980310471398e-05,
+      "loss": 16.1568,
+      "step": 3778
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.85387249801066e-05,
+      "loss": 17.9105,
+      "step": 3779
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.853764646353605e-05,
+      "loss": 16.2938,
+      "step": 3780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.853656755501104e-05,
+      "loss": 16.3478,
+      "step": 3781
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.853548825454029e-05,
+      "loss": 14.5969,
+      "step": 3782
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.853440856213249e-05,
+      "loss": 18.1936,
+      "step": 3783
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.85333284777964e-05,
+      "loss": 18.4734,
+      "step": 3784
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.85322480015407e-05,
+      "loss": 15.4231,
+      "step": 3785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.853116713337414e-05,
+      "loss": 16.0036,
+      "step": 3786
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.853008587330545e-05,
+      "loss": 18.8352,
+      "step": 3787
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.852900422134335e-05,
+      "loss": 16.6276,
+      "step": 3788
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.852792217749656e-05,
+      "loss": 16.997,
+      "step": 3789
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.852683974177386e-05,
+      "loss": 20.9698,
+      "step": 3790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.852575691418395e-05,
+      "loss": 18.5946,
+      "step": 3791
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.85246736947356e-05,
+      "loss": 17.3471,
+      "step": 3792
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.852359008343754e-05,
+      "loss": 16.4685,
+      "step": 3793
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.852250608029852e-05,
+      "loss": 17.496,
+      "step": 3794
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.852142168532732e-05,
+      "loss": 17.1234,
+      "step": 3795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.852033689853267e-05,
+      "loss": 16.4055,
+      "step": 3796
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.851925171992334e-05,
+      "loss": 16.8598,
+      "step": 3797
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.85181661495081e-05,
+      "loss": 16.8943,
+      "step": 3798
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.85170801872957e-05,
+      "loss": 16.6574,
+      "step": 3799
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.851599383329492e-05,
+      "loss": 17.2856,
+      "step": 3800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.851490708751453e-05,
+      "loss": 16.6939,
+      "step": 3801
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.851381994996331e-05,
+      "loss": 15.2038,
+      "step": 3802
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.851273242065002e-05,
+      "loss": 19.2148,
+      "step": 3803
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.851164449958348e-05,
+      "loss": 16.6151,
+      "step": 3804
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.851055618677242e-05,
+      "loss": 16.055,
+      "step": 3805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.850946748222569e-05,
+      "loss": 17.9812,
+      "step": 3806
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.850837838595203e-05,
+      "loss": 17.7977,
+      "step": 3807
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.850728889796028e-05,
+      "loss": 17.9522,
+      "step": 3808
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.850619901825921e-05,
+      "loss": 17.0874,
+      "step": 3809
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.850510874685761e-05,
+      "loss": 17.4703,
+      "step": 3810
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.850401808376432e-05,
+      "loss": 14.8519,
+      "step": 3811
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.85029270289881e-05,
+      "loss": 15.9681,
+      "step": 3812
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.850183558253782e-05,
+      "loss": 17.2105,
+      "step": 3813
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.850074374442226e-05,
+      "loss": 19.0517,
+      "step": 3814
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.849965151465022e-05,
+      "loss": 16.9271,
+      "step": 3815
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.849855889323054e-05,
+      "loss": 17.7652,
+      "step": 3816
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.849746588017206e-05,
+      "loss": 16.8955,
+      "step": 3817
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.849637247548356e-05,
+      "loss": 16.7073,
+      "step": 3818
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.849527867917392e-05,
+      "loss": 17.6427,
+      "step": 3819
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.849418449125194e-05,
+      "loss": 18.4835,
+      "step": 3820
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.849308991172648e-05,
+      "loss": 14.7001,
+      "step": 3821
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.849199494060635e-05,
+      "loss": 15.1067,
+      "step": 3822
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.849089957790041e-05,
+      "loss": 16.2794,
+      "step": 3823
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.848980382361752e-05,
+      "loss": 19.2036,
+      "step": 3824
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.84887076777665e-05,
+      "loss": 16.5068,
+      "step": 3825
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.848761114035621e-05,
+      "loss": 15.727,
+      "step": 3826
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.84865142113955e-05,
+      "loss": 19.1171,
+      "step": 3827
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.848541689089326e-05,
+      "loss": 17.0953,
+      "step": 3828
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.84843191788583e-05,
+      "loss": 17.8577,
+      "step": 3829
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.848322107529953e-05,
+      "loss": 16.2658,
+      "step": 3830
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.848212258022579e-05,
+      "loss": 14.7328,
+      "step": 3831
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.848102369364596e-05,
+      "loss": 17.672,
+      "step": 3832
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.847992441556893e-05,
+      "loss": 18.386,
+      "step": 3833
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.847882474600354e-05,
+      "loss": 16.0088,
+      "step": 3834
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.847772468495869e-05,
+      "loss": 17.9771,
+      "step": 3835
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.847662423244328e-05,
+      "loss": 16.0029,
+      "step": 3836
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.847552338846617e-05,
+      "loss": 19.4017,
+      "step": 3837
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.847442215303626e-05,
+      "loss": 16.6116,
+      "step": 3838
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.847332052616245e-05,
+      "loss": 15.3686,
+      "step": 3839
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.847221850785362e-05,
+      "loss": 16.9311,
+      "step": 3840
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.847111609811868e-05,
+      "loss": 15.8214,
+      "step": 3841
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.847001329696653e-05,
+      "loss": 16.597,
+      "step": 3842
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.846891010440607e-05,
+      "loss": 17.2352,
+      "step": 3843
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.846780652044623e-05,
+      "loss": 16.7992,
+      "step": 3844
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.846670254509589e-05,
+      "loss": 17.6392,
+      "step": 3845
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.846559817836398e-05,
+      "loss": 17.7772,
+      "step": 3846
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.846449342025943e-05,
+      "loss": 18.2521,
+      "step": 3847
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.846338827079114e-05,
+      "loss": 17.1689,
+      "step": 3848
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.846228272996805e-05,
+      "loss": 17.7459,
+      "step": 3849
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.846117679779908e-05,
+      "loss": 17.0424,
+      "step": 3850
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.846007047429315e-05,
+      "loss": 15.6718,
+      "step": 3851
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.845896375945923e-05,
+      "loss": 15.7916,
+      "step": 3852
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.845785665330623e-05,
+      "loss": 15.22,
+      "step": 3853
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.845674915584308e-05,
+      "loss": 15.2772,
+      "step": 3854
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.845564126707874e-05,
+      "loss": 15.4596,
+      "step": 3855
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.845453298702216e-05,
+      "loss": 19.5721,
+      "step": 3856
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.845342431568227e-05,
+      "loss": 18.7109,
+      "step": 3857
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.845231525306803e-05,
+      "loss": 17.6109,
+      "step": 3858
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.845120579918842e-05,
+      "loss": 18.6144,
+      "step": 3859
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.84500959540524e-05,
+      "loss": 19.8172,
+      "step": 3860
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.844898571766887e-05,
+      "loss": 17.4958,
+      "step": 3861
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.844787509004687e-05,
+      "loss": 16.537,
+      "step": 3862
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.844676407119532e-05,
+      "loss": 14.7262,
+      "step": 3863
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.844565266112322e-05,
+      "loss": 18.5437,
+      "step": 3864
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.844454085983952e-05,
+      "loss": 17.606,
+      "step": 3865
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.844342866735323e-05,
+      "loss": 16.5938,
+      "step": 3866
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.844231608367329e-05,
+      "loss": 16.6292,
+      "step": 3867
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.844120310880873e-05,
+      "loss": 18.2622,
+      "step": 3868
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.84400897427685e-05,
+      "loss": 15.9921,
+      "step": 3869
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.843897598556162e-05,
+      "loss": 16.4029,
+      "step": 3870
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.843786183719706e-05,
+      "loss": 18.2173,
+      "step": 3871
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.843674729768383e-05,
+      "loss": 17.231,
+      "step": 3872
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.843563236703093e-05,
+      "loss": 15.4139,
+      "step": 3873
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.843451704524736e-05,
+      "loss": 15.7287,
+      "step": 3874
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.843340133234213e-05,
+      "loss": 17.969,
+      "step": 3875
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.843228522832424e-05,
+      "loss": 15.23,
+      "step": 3876
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.843116873320271e-05,
+      "loss": 17.3066,
+      "step": 3877
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.843005184698655e-05,
+      "loss": 16.7719,
+      "step": 3878
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.84289345696848e-05,
+      "loss": 19.0944,
+      "step": 3879
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.842781690130644e-05,
+      "loss": 17.7999,
+      "step": 3880
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.842669884186056e-05,
+      "loss": 15.523,
+      "step": 3881
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.842558039135612e-05,
+      "loss": 14.8898,
+      "step": 3882
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.842446154980221e-05,
+      "loss": 16.8113,
+      "step": 3883
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.842334231720781e-05,
+      "loss": 17.3354,
+      "step": 3884
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.8422222693582e-05,
+      "loss": 18.5895,
+      "step": 3885
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.842110267893381e-05,
+      "loss": 15.741,
+      "step": 3886
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.841998227327228e-05,
+      "loss": 18.3244,
+      "step": 3887
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.841886147660645e-05,
+      "loss": 18.3625,
+      "step": 3888
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.84177402889454e-05,
+      "loss": 17.1402,
+      "step": 3889
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.841661871029814e-05,
+      "loss": 16.5657,
+      "step": 3890
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.841549674067377e-05,
+      "loss": 17.8517,
+      "step": 3891
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.841437438008131e-05,
+      "loss": 15.2781,
+      "step": 3892
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.841325162852986e-05,
+      "loss": 16.8907,
+      "step": 3893
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.841212848602846e-05,
+      "loss": 14.8903,
+      "step": 3894
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.84110049525862e-05,
+      "loss": 17.0561,
+      "step": 3895
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.840988102821214e-05,
+      "loss": 17.1882,
+      "step": 3896
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.840875671291536e-05,
+      "loss": 16.6591,
+      "step": 3897
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.840763200670495e-05,
+      "loss": 14.775,
+      "step": 3898
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.840650690958997e-05,
+      "loss": 15.621,
+      "step": 3899
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.840538142157952e-05,
+      "loss": 16.6669,
+      "step": 3900
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.840425554268267e-05,
+      "loss": 17.0382,
+      "step": 3901
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.840312927290854e-05,
+      "loss": 15.9374,
+      "step": 3902
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.840200261226622e-05,
+      "loss": 15.5041,
+      "step": 3903
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.840087556076478e-05,
+      "loss": 16.1914,
+      "step": 3904
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.839974811841335e-05,
+      "loss": 16.4371,
+      "step": 3905
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.839862028522103e-05,
+      "loss": 17.1087,
+      "step": 3906
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.839749206119692e-05,
+      "loss": 16.7427,
+      "step": 3907
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.839636344635012e-05,
+      "loss": 16.1077,
+      "step": 3908
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.839523444068978e-05,
+      "loss": 15.8577,
+      "step": 3909
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.8394105044225e-05,
+      "loss": 16.7657,
+      "step": 3910
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.839297525696487e-05,
+      "loss": 16.6723,
+      "step": 3911
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.839184507891855e-05,
+      "loss": 18.8237,
+      "step": 3912
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.839071451009514e-05,
+      "loss": 18.3766,
+      "step": 3913
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.83895835505038e-05,
+      "loss": 15.7108,
+      "step": 3914
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.838845220015365e-05,
+      "loss": 16.5562,
+      "step": 3915
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.838732045905382e-05,
+      "loss": 16.7613,
+      "step": 3916
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.838618832721345e-05,
+      "loss": 16.4532,
+      "step": 3917
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.838505580464168e-05,
+      "loss": 16.0173,
+      "step": 3918
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.838392289134766e-05,
+      "loss": 17.5529,
+      "step": 3919
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.838278958734055e-05,
+      "loss": 16.9002,
+      "step": 3920
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.838165589262948e-05,
+      "loss": 17.5509,
+      "step": 3921
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.838052180722361e-05,
+      "loss": 17.5279,
+      "step": 3922
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.837938733113211e-05,
+      "loss": 16.1034,
+      "step": 3923
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.837825246436414e-05,
+      "loss": 16.9261,
+      "step": 3924
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.837711720692885e-05,
+      "loss": 16.8522,
+      "step": 3925
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.837598155883541e-05,
+      "loss": 16.9373,
+      "step": 3926
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.837484552009302e-05,
+      "loss": 16.5721,
+      "step": 3927
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.83737090907108e-05,
+      "loss": 17.0751,
+      "step": 3928
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.837257227069798e-05,
+      "loss": 16.5476,
+      "step": 3929
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.837143506006371e-05,
+      "loss": 15.4274,
+      "step": 3930
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.837029745881719e-05,
+      "loss": 19.2411,
+      "step": 3931
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.836915946696759e-05,
+      "loss": 16.5141,
+      "step": 3932
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.836802108452409e-05,
+      "loss": 15.8004,
+      "step": 3933
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.836688231149592e-05,
+      "loss": 17.7326,
+      "step": 3934
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.836574314789225e-05,
+      "loss": 18.979,
+      "step": 3935
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.836460359372228e-05,
+      "loss": 16.4801,
+      "step": 3936
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.836346364899522e-05,
+      "loss": 17.0174,
+      "step": 3937
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.836232331372028e-05,
+      "loss": 17.4976,
+      "step": 3938
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.836118258790666e-05,
+      "loss": 18.0868,
+      "step": 3939
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.836004147156357e-05,
+      "loss": 16.989,
+      "step": 3940
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.835889996470022e-05,
+      "loss": 17.1327,
+      "step": 3941
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.835775806732584e-05,
+      "loss": 16.4788,
+      "step": 3942
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.835661577944965e-05,
+      "loss": 15.7376,
+      "step": 3943
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.835547310108087e-05,
+      "loss": 15.7951,
+      "step": 3944
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.835433003222872e-05,
+      "loss": 17.3552,
+      "step": 3945
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.835318657290244e-05,
+      "loss": 15.9278,
+      "step": 3946
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.835204272311127e-05,
+      "loss": 17.998,
+      "step": 3947
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.835089848286444e-05,
+      "loss": 15.6654,
+      "step": 3948
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.834975385217117e-05,
+      "loss": 17.6005,
+      "step": 3949
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.834860883104075e-05,
+      "loss": 17.519,
+      "step": 3950
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.834746341948238e-05,
+      "loss": 17.358,
+      "step": 3951
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.834631761750533e-05,
+      "loss": 18.257,
+      "step": 3952
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.834517142511886e-05,
+      "loss": 17.611,
+      "step": 3953
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.834402484233222e-05,
+      "loss": 16.2586,
+      "step": 3954
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.834287786915466e-05,
+      "loss": 16.0251,
+      "step": 3955
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.834173050559543e-05,
+      "loss": 17.5076,
+      "step": 3956
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.834058275166383e-05,
+      "loss": 16.5737,
+      "step": 3957
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.833943460736912e-05,
+      "loss": 14.744,
+      "step": 3958
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.833828607272056e-05,
+      "loss": 17.3718,
+      "step": 3959
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.833713714772741e-05,
+      "loss": 18.9754,
+      "step": 3960
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.833598783239897e-05,
+      "loss": 15.903,
+      "step": 3961
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.833483812674453e-05,
+      "loss": 18.2542,
+      "step": 3962
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.833368803077335e-05,
+      "loss": 17.1846,
+      "step": 3963
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.833253754449474e-05,
+      "loss": 17.1778,
+      "step": 3964
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.833138666791797e-05,
+      "loss": 16.7012,
+      "step": 3965
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.833023540105234e-05,
+      "loss": 16.3023,
+      "step": 3966
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.832908374390714e-05,
+      "loss": 16.6659,
+      "step": 3967
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.832793169649171e-05,
+      "loss": 18.8575,
+      "step": 3968
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.83267792588153e-05,
+      "loss": 15.5737,
+      "step": 3969
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.832562643088724e-05,
+      "loss": 17.2942,
+      "step": 3970
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.832447321271684e-05,
+      "loss": 16.1331,
+      "step": 3971
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.832331960431342e-05,
+      "loss": 15.5105,
+      "step": 3972
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.832216560568627e-05,
+      "loss": 15.7629,
+      "step": 3973
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.832101121684474e-05,
+      "loss": 15.762,
+      "step": 3974
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.831985643779811e-05,
+      "loss": 17.0341,
+      "step": 3975
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.831870126855576e-05,
+      "loss": 16.2163,
+      "step": 3976
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.831754570912699e-05,
+      "loss": 16.8017,
+      "step": 3977
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.831638975952111e-05,
+      "loss": 17.9501,
+      "step": 3978
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.831523341974749e-05,
+      "loss": 17.0446,
+      "step": 3979
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.831407668981546e-05,
+      "loss": 19.2189,
+      "step": 3980
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.831291956973434e-05,
+      "loss": 18.5997,
+      "step": 3981
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.831176205951352e-05,
+      "loss": 17.6664,
+      "step": 3982
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.831060415916227e-05,
+      "loss": 15.4341,
+      "step": 3983
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.830944586869002e-05,
+      "loss": 18.7878,
+      "step": 3984
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.830828718810607e-05,
+      "loss": 15.5164,
+      "step": 3985
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.830712811741982e-05,
+      "loss": 16.0342,
+      "step": 3986
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.830596865664059e-05,
+      "loss": 17.1877,
+      "step": 3987
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.830480880577778e-05,
+      "loss": 14.4192,
+      "step": 3988
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.83036485648407e-05,
+      "loss": 17.9476,
+      "step": 3989
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.830248793383878e-05,
+      "loss": 17.3699,
+      "step": 3990
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.830132691278137e-05,
+      "loss": 17.5435,
+      "step": 3991
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.830016550167782e-05,
+      "loss": 14.5912,
+      "step": 3992
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.829900370053755e-05,
+      "loss": 18.2685,
+      "step": 3993
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.829784150936993e-05,
+      "loss": 16.8724,
+      "step": 3994
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.829667892818432e-05,
+      "loss": 16.4617,
+      "step": 3995
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.829551595699014e-05,
+      "loss": 17.1297,
+      "step": 3996
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.829435259579676e-05,
+      "loss": 17.9404,
+      "step": 3997
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.829318884461359e-05,
+      "loss": 16.0225,
+      "step": 3998
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.829202470345002e-05,
+      "loss": 17.4848,
+      "step": 3999
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.829086017231544e-05,
+      "loss": 16.9816,
+      "step": 4000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.828969525121927e-05,
+      "loss": 17.2665,
+      "step": 4001
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.828852994017092e-05,
+      "loss": 16.6303,
+      "step": 4002
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.828736423917978e-05,
+      "loss": 15.857,
+      "step": 4003
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.828619814825528e-05,
+      "loss": 16.8622,
+      "step": 4004
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.828503166740684e-05,
+      "loss": 18.8256,
+      "step": 4005
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.828386479664385e-05,
+      "loss": 16.2966,
+      "step": 4006
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.828269753597576e-05,
+      "loss": 16.4348,
+      "step": 4007
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.828152988541201e-05,
+      "loss": 14.9013,
+      "step": 4008
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.8280361844962e-05,
+      "loss": 16.0328,
+      "step": 4009
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.827919341463516e-05,
+      "loss": 18.4423,
+      "step": 4010
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.827802459444096e-05,
+      "loss": 16.5281,
+      "step": 4011
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.827685538438879e-05,
+      "loss": 13.3637,
+      "step": 4012
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.827568578448812e-05,
+      "loss": 17.7311,
+      "step": 4013
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.82745157947484e-05,
+      "loss": 18.0015,
+      "step": 4014
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.827334541517907e-05,
+      "loss": 18.6341,
+      "step": 4015
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.827217464578958e-05,
+      "loss": 17.9032,
+      "step": 4016
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.82710034865894e-05,
+      "loss": 15.6623,
+      "step": 4017
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.826983193758794e-05,
+      "loss": 18.9534,
+      "step": 4018
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.826865999879473e-05,
+      "loss": 16.5539,
+      "step": 4019
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.826748767021916e-05,
+      "loss": 15.5907,
+      "step": 4020
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.826631495187075e-05,
+      "loss": 15.0822,
+      "step": 4021
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.826514184375897e-05,
+      "loss": 14.8439,
+      "step": 4022
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.826396834589326e-05,
+      "loss": 16.0395,
+      "step": 4023
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.826279445828311e-05,
+      "loss": 16.3547,
+      "step": 4024
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.8261620180938e-05,
+      "loss": 16.5342,
+      "step": 4025
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.826044551386744e-05,
+      "loss": 17.6027,
+      "step": 4026
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.825927045708088e-05,
+      "loss": 17.1542,
+      "step": 4027
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.82580950105878e-05,
+      "loss": 19.6433,
+      "step": 4028
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.825691917439774e-05,
+      "loss": 17.2301,
+      "step": 4029
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.825574294852016e-05,
+      "loss": 18.3205,
+      "step": 4030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.825456633296454e-05,
+      "loss": 17.1332,
+      "step": 4031
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.825338932774044e-05,
+      "loss": 17.303,
+      "step": 4032
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.825221193285733e-05,
+      "loss": 17.5704,
+      "step": 4033
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.82510341483247e-05,
+      "loss": 17.7722,
+      "step": 4034
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.824985597415209e-05,
+      "loss": 17.293,
+      "step": 4035
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.8248677410349e-05,
+      "loss": 19.8072,
+      "step": 4036
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.824749845692496e-05,
+      "loss": 15.1014,
+      "step": 4037
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.824631911388948e-05,
+      "loss": 17.6916,
+      "step": 4038
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.824513938125208e-05,
+      "loss": 16.0874,
+      "step": 4039
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.82439592590223e-05,
+      "loss": 17.6511,
+      "step": 4040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.824277874720966e-05,
+      "loss": 18.4727,
+      "step": 4041
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.824159784582368e-05,
+      "loss": 16.112,
+      "step": 4042
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.824041655487392e-05,
+      "loss": 16.5187,
+      "step": 4043
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.823923487436992e-05,
+      "loss": 17.2544,
+      "step": 4044
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.823805280432121e-05,
+      "loss": 19.15,
+      "step": 4045
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.823687034473735e-05,
+      "loss": 16.3407,
+      "step": 4046
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.823568749562785e-05,
+      "loss": 15.3664,
+      "step": 4047
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.82345042570023e-05,
+      "loss": 17.6976,
+      "step": 4048
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.823332062887025e-05,
+      "loss": 15.7212,
+      "step": 4049
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.823213661124124e-05,
+      "loss": 17.6114,
+      "step": 4050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.823095220412486e-05,
+      "loss": 16.2185,
+      "step": 4051
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.822976740753064e-05,
+      "loss": 18.9552,
+      "step": 4052
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.822858222146816e-05,
+      "loss": 15.6895,
+      "step": 4053
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.822739664594702e-05,
+      "loss": 20.5498,
+      "step": 4054
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.822621068097675e-05,
+      "loss": 15.3564,
+      "step": 4055
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.822502432656697e-05,
+      "loss": 17.413,
+      "step": 4056
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.822383758272721e-05,
+      "loss": 17.603,
+      "step": 4057
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.82226504494671e-05,
+      "loss": 18.5898,
+      "step": 4058
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.822146292679619e-05,
+      "loss": 16.5735,
+      "step": 4059
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.822027501472409e-05,
+      "loss": 16.0377,
+      "step": 4060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.82190867132604e-05,
+      "loss": 16.927,
+      "step": 4061
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.821789802241469e-05,
+      "loss": 18.0532,
+      "step": 4062
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.821670894219658e-05,
+      "loss": 17.9467,
+      "step": 4063
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.821551947261567e-05,
+      "loss": 16.0871,
+      "step": 4064
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.821432961368156e-05,
+      "loss": 18.407,
+      "step": 4065
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.821313936540385e-05,
+      "loss": 16.7276,
+      "step": 4066
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.821194872779215e-05,
+      "loss": 17.4456,
+      "step": 4067
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.821075770085612e-05,
+      "loss": 15.4218,
+      "step": 4068
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.820956628460532e-05,
+      "loss": 17.6607,
+      "step": 4069
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.82083744790494e-05,
+      "loss": 16.8892,
+      "step": 4070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.820718228419798e-05,
+      "loss": 16.1605,
+      "step": 4071
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.820598970006069e-05,
+      "loss": 16.8499,
+      "step": 4072
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.820479672664715e-05,
+      "loss": 17.4254,
+      "step": 4073
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.8203603363967e-05,
+      "loss": 15.6441,
+      "step": 4074
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.820240961202987e-05,
+      "loss": 18.2944,
+      "step": 4075
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.820121547084542e-05,
+      "loss": 16.6075,
+      "step": 4076
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.820002094042328e-05,
+      "loss": 16.5771,
+      "step": 4077
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.819882602077309e-05,
+      "loss": 16.7907,
+      "step": 4078
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.819763071190449e-05,
+      "loss": 14.2356,
+      "step": 4079
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.819643501382717e-05,
+      "loss": 16.5095,
+      "step": 4080
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.819523892655076e-05,
+      "loss": 16.6433,
+      "step": 4081
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.819404245008493e-05,
+      "loss": 16.1121,
+      "step": 4082
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.819284558443931e-05,
+      "loss": 16.4496,
+      "step": 4083
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.81916483296236e-05,
+      "loss": 15.9445,
+      "step": 4084
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.819045068564745e-05,
+      "loss": 16.4942,
+      "step": 4085
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.818925265252055e-05,
+      "loss": 16.4996,
+      "step": 4086
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.818805423025255e-05,
+      "loss": 18.5606,
+      "step": 4087
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.818685541885315e-05,
+      "loss": 15.9685,
+      "step": 4088
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.818565621833203e-05,
+      "loss": 16.9292,
+      "step": 4089
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.818445662869884e-05,
+      "loss": 16.1985,
+      "step": 4090
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.81832566499633e-05,
+      "loss": 16.3106,
+      "step": 4091
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.818205628213511e-05,
+      "loss": 17.7814,
+      "step": 4092
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.818085552522392e-05,
+      "loss": 16.5825,
+      "step": 4093
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.817965437923946e-05,
+      "loss": 16.9982,
+      "step": 4094
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.817845284419142e-05,
+      "loss": 16.8656,
+      "step": 4095
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.817725092008951e-05,
+      "loss": 17.0875,
+      "step": 4096
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.817604860694342e-05,
+      "loss": 17.8979,
+      "step": 4097
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.817484590476289e-05,
+      "loss": 15.8916,
+      "step": 4098
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.817364281355759e-05,
+      "loss": 18.5031,
+      "step": 4099
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.817243933333727e-05,
+      "loss": 17.2094,
+      "step": 4100
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.817123546411162e-05,
+      "loss": 16.1736,
+      "step": 4101
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.817003120589039e-05,
+      "loss": 18.8043,
+      "step": 4102
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.816882655868329e-05,
+      "loss": 17.4027,
+      "step": 4103
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.816762152250003e-05,
+      "loss": 18.7281,
+      "step": 4104
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.816641609735038e-05,
+      "loss": 17.3882,
+      "step": 4105
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.816521028324403e-05,
+      "loss": 16.1935,
+      "step": 4106
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.816400408019078e-05,
+      "loss": 17.4604,
+      "step": 4107
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.81627974882003e-05,
+      "loss": 16.8088,
+      "step": 4108
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.816159050728238e-05,
+      "loss": 18.4636,
+      "step": 4109
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.816038313744673e-05,
+      "loss": 17.9489,
+      "step": 4110
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.815917537870314e-05,
+      "loss": 18.0325,
+      "step": 4111
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.815796723106135e-05,
+      "loss": 16.609,
+      "step": 4112
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.815675869453111e-05,
+      "loss": 16.2873,
+      "step": 4113
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.815554976912215e-05,
+      "loss": 16.9367,
+      "step": 4114
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.815434045484431e-05,
+      "loss": 16.9304,
+      "step": 4115
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.815313075170727e-05,
+      "loss": 15.5644,
+      "step": 4116
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.815192065972085e-05,
+      "loss": 15.7378,
+      "step": 4117
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.815071017889482e-05,
+      "loss": 17.4707,
+      "step": 4118
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.814949930923892e-05,
+      "loss": 17.7079,
+      "step": 4119
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.814828805076297e-05,
+      "loss": 17.053,
+      "step": 4120
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.81470764034767e-05,
+      "loss": 19.3687,
+      "step": 4121
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.814586436738998e-05,
+      "loss": 16.6957,
+      "step": 4122
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.81446519425125e-05,
+      "loss": 14.7385,
+      "step": 4123
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.814343912885411e-05,
+      "loss": 18.2001,
+      "step": 4124
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.814222592642459e-05,
+      "loss": 18.776,
+      "step": 4125
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.814101233523374e-05,
+      "loss": 17.2723,
+      "step": 4126
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.813979835529136e-05,
+      "loss": 20.0151,
+      "step": 4127
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.813858398660724e-05,
+      "loss": 17.2916,
+      "step": 4128
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.813736922919119e-05,
+      "loss": 16.9444,
+      "step": 4129
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.813615408305304e-05,
+      "loss": 17.6131,
+      "step": 4130
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.813493854820257e-05,
+      "loss": 15.8811,
+      "step": 4131
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.813372262464964e-05,
+      "loss": 17.1586,
+      "step": 4132
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.813250631240402e-05,
+      "loss": 17.973,
+      "step": 4133
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.813128961147555e-05,
+      "loss": 18.0064,
+      "step": 4134
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.813007252187407e-05,
+      "loss": 16.951,
+      "step": 4135
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.812885504360941e-05,
+      "loss": 15.5964,
+      "step": 4136
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.812763717669136e-05,
+      "loss": 16.3118,
+      "step": 4137
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.812641892112982e-05,
+      "loss": 16.0979,
+      "step": 4138
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.812520027693458e-05,
+      "loss": 15.8748,
+      "step": 4139
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.812398124411547e-05,
+      "loss": 16.3182,
+      "step": 4140
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.812276182268236e-05,
+      "loss": 16.4333,
+      "step": 4141
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.812154201264511e-05,
+      "loss": 16.9713,
+      "step": 4142
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.812032181401356e-05,
+      "loss": 17.3119,
+      "step": 4143
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.811910122679753e-05,
+      "loss": 16.6149,
+      "step": 4144
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.811788025100693e-05,
+      "loss": 16.3332,
+      "step": 4145
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.811665888665159e-05,
+      "loss": 15.1744,
+      "step": 4146
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.811543713374136e-05,
+      "loss": 18.5638,
+      "step": 4147
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.811421499228614e-05,
+      "loss": 16.3511,
+      "step": 4148
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.811299246229578e-05,
+      "loss": 17.5539,
+      "step": 4149
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.811176954378014e-05,
+      "loss": 16.124,
+      "step": 4150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.811054623674912e-05,
+      "loss": 18.1823,
+      "step": 4151
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.810932254121259e-05,
+      "loss": 19.5029,
+      "step": 4152
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.810809845718043e-05,
+      "loss": 16.6774,
+      "step": 4153
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.810687398466251e-05,
+      "loss": 17.3283,
+      "step": 4154
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.810564912366875e-05,
+      "loss": 15.6159,
+      "step": 4155
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.810442387420902e-05,
+      "loss": 16.6479,
+      "step": 4156
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.810319823629322e-05,
+      "loss": 16.2417,
+      "step": 4157
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.810197220993123e-05,
+      "loss": 17.2612,
+      "step": 4158
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.810074579513298e-05,
+      "loss": 16.2748,
+      "step": 4159
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.809951899190835e-05,
+      "loss": 16.6067,
+      "step": 4160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.809829180026725e-05,
+      "loss": 16.4669,
+      "step": 4161
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80970642202196e-05,
+      "loss": 16.667,
+      "step": 4162
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.809583625177532e-05,
+      "loss": 16.3613,
+      "step": 4163
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.809460789494432e-05,
+      "loss": 16.4527,
+      "step": 4164
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.809337914973648e-05,
+      "loss": 17.337,
+      "step": 4165
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.809215001616179e-05,
+      "loss": 16.4959,
+      "step": 4166
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.809092049423012e-05,
+      "loss": 17.4914,
+      "step": 4167
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.808969058395144e-05,
+      "loss": 15.9268,
+      "step": 4168
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.808846028533564e-05,
+      "loss": 17.6486,
+      "step": 4169
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80872295983927e-05,
+      "loss": 15.9408,
+      "step": 4170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.808599852313254e-05,
+      "loss": 17.1062,
+      "step": 4171
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.808476705956508e-05,
+      "loss": 17.124,
+      "step": 4172
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.808353520770028e-05,
+      "loss": 14.1375,
+      "step": 4173
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80823029675481e-05,
+      "loss": 17.8528,
+      "step": 4174
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.808107033911848e-05,
+      "loss": 15.7129,
+      "step": 4175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.807983732242137e-05,
+      "loss": 17.326,
+      "step": 4176
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.807860391746673e-05,
+      "loss": 16.8967,
+      "step": 4177
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.807737012426452e-05,
+      "loss": 14.8843,
+      "step": 4178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.807613594282469e-05,
+      "loss": 17.0701,
+      "step": 4179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.807490137315725e-05,
+      "loss": 17.515,
+      "step": 4180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.807366641527212e-05,
+      "loss": 18.4336,
+      "step": 4181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80724310691793e-05,
+      "loss": 14.1285,
+      "step": 4182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.807119533488876e-05,
+      "loss": 17.3664,
+      "step": 4183
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.806995921241047e-05,
+      "loss": 15.7567,
+      "step": 4184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.806872270175442e-05,
+      "loss": 16.6363,
+      "step": 4185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.806748580293059e-05,
+      "loss": 18.2434,
+      "step": 4186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.806624851594898e-05,
+      "loss": 18.1454,
+      "step": 4187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.806501084081957e-05,
+      "loss": 17.2911,
+      "step": 4188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.806377277755237e-05,
+      "loss": 16.7706,
+      "step": 4189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.806253432615735e-05,
+      "loss": 17.5023,
+      "step": 4190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.806129548664453e-05,
+      "loss": 17.5243,
+      "step": 4191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.806005625902391e-05,
+      "loss": 17.5959,
+      "step": 4192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80588166433055e-05,
+      "loss": 18.0688,
+      "step": 4193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80575766394993e-05,
+      "loss": 18.8726,
+      "step": 4194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.805633624761536e-05,
+      "loss": 17.5817,
+      "step": 4195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.805509546766364e-05,
+      "loss": 16.7542,
+      "step": 4196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.805385429965419e-05,
+      "loss": 14.6693,
+      "step": 4197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.805261274359705e-05,
+      "loss": 16.8666,
+      "step": 4198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80513707995022e-05,
+      "loss": 17.7879,
+      "step": 4199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80501284673797e-05,
+      "loss": 17.1414,
+      "step": 4200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80488857472396e-05,
+      "loss": 16.909,
+      "step": 4201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80476426390919e-05,
+      "loss": 18.5873,
+      "step": 4202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.804639914294664e-05,
+      "loss": 16.1466,
+      "step": 4203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.804515525881388e-05,
+      "loss": 16.0193,
+      "step": 4204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.804391098670367e-05,
+      "loss": 17.0262,
+      "step": 4205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.804266632662603e-05,
+      "loss": 16.9598,
+      "step": 4206
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.804142127859104e-05,
+      "loss": 15.1859,
+      "step": 4207
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.804017584260873e-05,
+      "loss": 15.2274,
+      "step": 4208
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.803893001868916e-05,
+      "loss": 15.9399,
+      "step": 4209
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.803768380684242e-05,
+      "loss": 17.8657,
+      "step": 4210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.803643720707853e-05,
+      "loss": 17.5541,
+      "step": 4211
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.803519021940759e-05,
+      "loss": 17.2392,
+      "step": 4212
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.803394284383965e-05,
+      "loss": 14.6507,
+      "step": 4213
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80326950803848e-05,
+      "loss": 17.499,
+      "step": 4214
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80314469290531e-05,
+      "loss": 15.7557,
+      "step": 4215
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.803019838985463e-05,
+      "loss": 16.8521,
+      "step": 4216
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80289494627995e-05,
+      "loss": 17.2996,
+      "step": 4217
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.802770014789775e-05,
+      "loss": 15.1779,
+      "step": 4218
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80264504451595e-05,
+      "loss": 17.7475,
+      "step": 4219
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.802520035459484e-05,
+      "loss": 17.1374,
+      "step": 4220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.802394987621387e-05,
+      "loss": 15.9799,
+      "step": 4221
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.802269901002665e-05,
+      "loss": 17.371,
+      "step": 4222
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.802144775604332e-05,
+      "loss": 17.14,
+      "step": 4223
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.802019611427396e-05,
+      "loss": 18.1741,
+      "step": 4224
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80189440847287e-05,
+      "loss": 16.9631,
+      "step": 4225
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.801769166741764e-05,
+      "loss": 17.8649,
+      "step": 4226
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.80164388623509e-05,
+      "loss": 17.4094,
+      "step": 4227
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.801518566953858e-05,
+      "loss": 18.0716,
+      "step": 4228
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.801393208899081e-05,
+      "loss": 17.8614,
+      "step": 4229
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.801267812071771e-05,
+      "loss": 16.2983,
+      "step": 4230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.801142376472942e-05,
+      "loss": 18.4829,
+      "step": 4231
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.801016902103606e-05,
+      "loss": 15.0798,
+      "step": 4232
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.800891388964774e-05,
+      "loss": 16.8703,
+      "step": 4233
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.800765837057464e-05,
+      "loss": 16.1148,
+      "step": 4234
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.800640246382687e-05,
+      "loss": 14.6829,
+      "step": 4235
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.800514616941457e-05,
+      "loss": 16.1994,
+      "step": 4236
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.800388948734789e-05,
+      "loss": 15.4606,
+      "step": 4237
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.800263241763698e-05,
+      "loss": 17.6357,
+      "step": 4238
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.8001374960292e-05,
+      "loss": 15.2281,
+      "step": 4239
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.800011711532308e-05,
+      "loss": 15.7551,
+      "step": 4240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.79988588827404e-05,
+      "loss": 14.7408,
+      "step": 4241
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.799760026255412e-05,
+      "loss": 17.1841,
+      "step": 4242
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.799634125477439e-05,
+      "loss": 17.2633,
+      "step": 4243
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.799508185941138e-05,
+      "loss": 16.8765,
+      "step": 4244
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.799382207647526e-05,
+      "loss": 18.8682,
+      "step": 4245
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.799256190597621e-05,
+      "loss": 17.992,
+      "step": 4246
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.799130134792442e-05,
+      "loss": 16.2719,
+      "step": 4247
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.799004040233004e-05,
+      "loss": 16.2975,
+      "step": 4248
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.798877906920326e-05,
+      "loss": 17.3048,
+      "step": 4249
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.798751734855426e-05,
+      "loss": 15.5972,
+      "step": 4250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.798625524039327e-05,
+      "loss": 16.8626,
+      "step": 4251
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.798499274473043e-05,
+      "loss": 18.6346,
+      "step": 4252
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.798372986157596e-05,
+      "loss": 16.5944,
+      "step": 4253
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.798246659094006e-05,
+      "loss": 16.9549,
+      "step": 4254
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.79812029328329e-05,
+      "loss": 15.6931,
+      "step": 4255
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.797993888726473e-05,
+      "loss": 17.983,
+      "step": 4256
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.797867445424575e-05,
+      "loss": 18.4972,
+      "step": 4257
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.797740963378615e-05,
+      "loss": 16.7997,
+      "step": 4258
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.797614442589613e-05,
+      "loss": 16.9214,
+      "step": 4259
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.797487883058595e-05,
+      "loss": 17.2908,
+      "step": 4260
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.797361284786579e-05,
+      "loss": 17.9006,
+      "step": 4261
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.797234647774591e-05,
+      "loss": 16.3149,
+      "step": 4262
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.797107972023652e-05,
+      "loss": 17.2282,
+      "step": 4263
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.796981257534783e-05,
+      "loss": 18.4434,
+      "step": 4264
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.79685450430901e-05,
+      "loss": 17.4915,
+      "step": 4265
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.796727712347356e-05,
+      "loss": 17.0116,
+      "step": 4266
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.796600881650844e-05,
+      "loss": 16.6581,
+      "step": 4267
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.796474012220498e-05,
+      "loss": 14.8035,
+      "step": 4268
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.796347104057343e-05,
+      "loss": 16.7345,
+      "step": 4269
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.796220157162404e-05,
+      "loss": 18.3673,
+      "step": 4270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.796093171536707e-05,
+      "loss": 17.6026,
+      "step": 4271
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.795966147181276e-05,
+      "loss": 15.5933,
+      "step": 4272
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.795839084097138e-05,
+      "loss": 16.2216,
+      "step": 4273
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.795711982285316e-05,
+      "loss": 15.7185,
+      "step": 4274
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.795584841746841e-05,
+      "loss": 15.2457,
+      "step": 4275
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.795457662482737e-05,
+      "loss": 16.0615,
+      "step": 4276
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.79533044449403e-05,
+      "loss": 16.5828,
+      "step": 4277
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.795203187781751e-05,
+      "loss": 18.0114,
+      "step": 4278
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.795075892346923e-05,
+      "loss": 15.7938,
+      "step": 4279
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.794948558190577e-05,
+      "loss": 16.5619,
+      "step": 4280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.794821185313743e-05,
+      "loss": 17.2481,
+      "step": 4281
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.794693773717445e-05,
+      "loss": 16.2435,
+      "step": 4282
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.794566323402714e-05,
+      "loss": 17.943,
+      "step": 4283
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.79443883437058e-05,
+      "loss": 15.838,
+      "step": 4284
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.79431130662207e-05,
+      "loss": 15.617,
+      "step": 4285
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.794183740158218e-05,
+      "loss": 16.7826,
+      "step": 4286
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.794056134980049e-05,
+      "loss": 16.262,
+      "step": 4287
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.793928491088598e-05,
+      "loss": 17.963,
+      "step": 4288
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.793800808484895e-05,
+      "loss": 18.1154,
+      "step": 4289
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.793673087169967e-05,
+      "loss": 16.9182,
+      "step": 4290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.79354532714485e-05,
+      "loss": 18.8605,
+      "step": 4291
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.793417528410574e-05,
+      "loss": 17.455,
+      "step": 4292
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.793289690968171e-05,
+      "loss": 16.8905,
+      "step": 4293
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.793161814818673e-05,
+      "loss": 16.2606,
+      "step": 4294
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.793033899963112e-05,
+      "loss": 16.362,
+      "step": 4295
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.792905946402525e-05,
+      "loss": 16.3957,
+      "step": 4296
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.792777954137939e-05,
+      "loss": 18.9339,
+      "step": 4297
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.792649923170392e-05,
+      "loss": 17.3423,
+      "step": 4298
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.792521853500916e-05,
+      "loss": 16.6067,
+      "step": 4299
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.792393745130547e-05,
+      "loss": 17.8759,
+      "step": 4300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.792265598060317e-05,
+      "loss": 16.3134,
+      "step": 4301
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.792137412291265e-05,
+      "loss": 16.9731,
+      "step": 4302
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.79200918782442e-05,
+      "loss": 16.7711,
+      "step": 4303
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.791880924660823e-05,
+      "loss": 15.8755,
+      "step": 4304
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.791752622801507e-05,
+      "loss": 17.7577,
+      "step": 4305
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.791624282247507e-05,
+      "loss": 16.8905,
+      "step": 4306
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.791495902999863e-05,
+      "loss": 15.6354,
+      "step": 4307
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.791367485059608e-05,
+      "loss": 16.3946,
+      "step": 4308
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.791239028427783e-05,
+      "loss": 16.4185,
+      "step": 4309
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.791110533105421e-05,
+      "loss": 17.2083,
+      "step": 4310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.790981999093563e-05,
+      "loss": 17.1858,
+      "step": 4311
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.790853426393245e-05,
+      "loss": 17.2108,
+      "step": 4312
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.790724815005505e-05,
+      "loss": 15.0542,
+      "step": 4313
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.790596164931383e-05,
+      "loss": 15.0036,
+      "step": 4314
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.790467476171919e-05,
+      "loss": 14.0405,
+      "step": 4315
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.790338748728148e-05,
+      "loss": 17.3445,
+      "step": 4316
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.790209982601113e-05,
+      "loss": 16.388,
+      "step": 4317
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.790081177791852e-05,
+      "loss": 15.4305,
+      "step": 4318
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.789952334301407e-05,
+      "loss": 17.3691,
+      "step": 4319
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.789823452130818e-05,
+      "loss": 14.0893,
+      "step": 4320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.789694531281124e-05,
+      "loss": 17.9661,
+      "step": 4321
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.789565571753368e-05,
+      "loss": 17.9568,
+      "step": 4322
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.789436573548589e-05,
+      "loss": 17.5774,
+      "step": 4323
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.789307536667832e-05,
+      "loss": 17.3496,
+      "step": 4324
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.789178461112137e-05,
+      "loss": 15.0453,
+      "step": 4325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.789049346882547e-05,
+      "loss": 13.7456,
+      "step": 4326
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.788920193980103e-05,
+      "loss": 18.3354,
+      "step": 4327
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.78879100240585e-05,
+      "loss": 15.7964,
+      "step": 4328
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.788661772160831e-05,
+      "loss": 17.567,
+      "step": 4329
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.788532503246088e-05,
+      "loss": 19.5848,
+      "step": 4330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.788403195662668e-05,
+      "loss": 16.152,
+      "step": 4331
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.788273849411612e-05,
+      "loss": 17.2691,
+      "step": 4332
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.788144464493964e-05,
+      "loss": 17.101,
+      "step": 4333
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.788015040910771e-05,
+      "loss": 17.747,
+      "step": 4334
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.787885578663078e-05,
+      "loss": 17.6086,
+      "step": 4335
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.787756077751931e-05,
+      "loss": 15.2195,
+      "step": 4336
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.787626538178374e-05,
+      "loss": 18.0232,
+      "step": 4337
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.787496959943453e-05,
+      "loss": 17.7403,
+      "step": 4338
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.787367343048216e-05,
+      "loss": 17.7546,
+      "step": 4339
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.787237687493709e-05,
+      "loss": 17.4894,
+      "step": 4340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.78710799328098e-05,
+      "loss": 17.5015,
+      "step": 4341
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.786978260411073e-05,
+      "loss": 19.2689,
+      "step": 4342
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.786848488885039e-05,
+      "loss": 16.4025,
+      "step": 4343
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.786718678703925e-05,
+      "loss": 16.443,
+      "step": 4344
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.786588829868778e-05,
+      "loss": 16.0343,
+      "step": 4345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.786458942380649e-05,
+      "loss": 16.5045,
+      "step": 4346
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.786329016240584e-05,
+      "loss": 16.5357,
+      "step": 4347
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.786199051449636e-05,
+      "loss": 17.0384,
+      "step": 4348
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.78606904800885e-05,
+      "loss": 16.1815,
+      "step": 4349
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.785939005919278e-05,
+      "loss": 15.2135,
+      "step": 4350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.78580892518197e-05,
+      "loss": 18.4617,
+      "step": 4351
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.785678805797977e-05,
+      "loss": 15.0978,
+      "step": 4352
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.78554864776835e-05,
+      "loss": 16.5185,
+      "step": 4353
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.78541845109414e-05,
+      "loss": 16.6439,
+      "step": 4354
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.785288215776396e-05,
+      "loss": 16.6,
+      "step": 4355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.785157941816172e-05,
+      "loss": 18.2144,
+      "step": 4356
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.78502762921452e-05,
+      "loss": 18.6421,
+      "step": 4357
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.784897277972491e-05,
+      "loss": 17.872,
+      "step": 4358
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.784766888091139e-05,
+      "loss": 14.9047,
+      "step": 4359
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.784636459571515e-05,
+      "loss": 15.9658,
+      "step": 4360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.784505992414674e-05,
+      "loss": 16.6244,
+      "step": 4361
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.78437548662167e-05,
+      "loss": 16.2904,
+      "step": 4362
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.784244942193554e-05,
+      "loss": 16.8824,
+      "step": 4363
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.784114359131385e-05,
+      "loss": 15.8816,
+      "step": 4364
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.783983737436213e-05,
+      "loss": 18.5748,
+      "step": 4365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.783853077109095e-05,
+      "loss": 17.3525,
+      "step": 4366
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.783722378151085e-05,
+      "loss": 16.349,
+      "step": 4367
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.783591640563238e-05,
+      "loss": 16.2696,
+      "step": 4368
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.783460864346612e-05,
+      "loss": 14.8448,
+      "step": 4369
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.783330049502262e-05,
+      "loss": 18.5255,
+      "step": 4370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.783199196031243e-05,
+      "loss": 16.7875,
+      "step": 4371
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.783068303934613e-05,
+      "loss": 17.5094,
+      "step": 4372
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.78293737321343e-05,
+      "loss": 18.2632,
+      "step": 4373
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.782806403868749e-05,
+      "loss": 14.9376,
+      "step": 4374
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.782675395901628e-05,
+      "loss": 17.2552,
+      "step": 4375
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.782544349313128e-05,
+      "loss": 15.8599,
+      "step": 4376
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.782413264104304e-05,
+      "loss": 18.1122,
+      "step": 4377
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.782282140276213e-05,
+      "loss": 18.5421,
+      "step": 4378
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.782150977829919e-05,
+      "loss": 18.3604,
+      "step": 4379
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.782019776766478e-05,
+      "loss": 17.2936,
+      "step": 4380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.781888537086949e-05,
+      "loss": 18.473,
+      "step": 4381
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.781757258792394e-05,
+      "loss": 16.3386,
+      "step": 4382
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.78162594188387e-05,
+      "loss": 16.4327,
+      "step": 4383
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.781494586362441e-05,
+      "loss": 17.7113,
+      "step": 4384
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.781363192229165e-05,
+      "loss": 19.4694,
+      "step": 4385
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.781231759485106e-05,
+      "loss": 16.2055,
+      "step": 4386
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.78110028813132e-05,
+      "loss": 16.9651,
+      "step": 4387
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.780968778168874e-05,
+      "loss": 17.6322,
+      "step": 4388
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.780837229598828e-05,
+      "loss": 17.0466,
+      "step": 4389
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.780705642422244e-05,
+      "loss": 17.3329,
+      "step": 4390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.780574016640185e-05,
+      "loss": 16.5891,
+      "step": 4391
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.780442352253714e-05,
+      "loss": 15.8498,
+      "step": 4392
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.780310649263894e-05,
+      "loss": 18.5196,
+      "step": 4393
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.780178907671789e-05,
+      "loss": 15.6117,
+      "step": 4394
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.780047127478462e-05,
+      "loss": 17.4373,
+      "step": 4395
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.779915308684977e-05,
+      "loss": 16.2729,
+      "step": 4396
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.779783451292399e-05,
+      "loss": 17.138,
+      "step": 4397
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.779651555301794e-05,
+      "loss": 16.6642,
+      "step": 4398
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.779519620714226e-05,
+      "loss": 16.0172,
+      "step": 4399
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.77938764753076e-05,
+      "loss": 17.5164,
+      "step": 4400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.779255635752463e-05,
+      "loss": 17.3778,
+      "step": 4401
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.779123585380398e-05,
+      "loss": 17.238,
+      "step": 4402
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.778991496415634e-05,
+      "loss": 17.2789,
+      "step": 4403
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.778859368859237e-05,
+      "loss": 17.5506,
+      "step": 4404
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.778727202712274e-05,
+      "loss": 17.7027,
+      "step": 4405
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.778594997975813e-05,
+      "loss": 17.5002,
+      "step": 4406
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.778462754650921e-05,
+      "loss": 17.3606,
+      "step": 4407
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.778330472738666e-05,
+      "loss": 15.6958,
+      "step": 4408
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.778198152240116e-05,
+      "loss": 17.0227,
+      "step": 4409
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.778065793156339e-05,
+      "loss": 17.108,
+      "step": 4410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.777933395488405e-05,
+      "loss": 16.1703,
+      "step": 4411
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.777800959237382e-05,
+      "loss": 18.6405,
+      "step": 4412
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.777668484404341e-05,
+      "loss": 16.8279,
+      "step": 4413
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.77753597099035e-05,
+      "loss": 17.0315,
+      "step": 4414
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.77740341899648e-05,
+      "loss": 17.666,
+      "step": 4415
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.777270828423801e-05,
+      "loss": 16.7362,
+      "step": 4416
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.777138199273384e-05,
+      "loss": 16.7021,
+      "step": 4417
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.7770055315463e-05,
+      "loss": 16.8114,
+      "step": 4418
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.776872825243621e-05,
+      "loss": 15.4122,
+      "step": 4419
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.776740080366417e-05,
+      "loss": 14.7497,
+      "step": 4420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.776607296915761e-05,
+      "loss": 15.7182,
+      "step": 4421
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.776474474892725e-05,
+      "loss": 18.1318,
+      "step": 4422
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.776341614298383e-05,
+      "loss": 15.5598,
+      "step": 4423
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.776208715133806e-05,
+      "loss": 16.5724,
+      "step": 4424
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.776075777400067e-05,
+      "loss": 17.6174,
+      "step": 4425
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.77594280109824e-05,
+      "loss": 14.8462,
+      "step": 4426
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.7758097862294e-05,
+      "loss": 18.4737,
+      "step": 4427
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.775676732794621e-05,
+      "loss": 16.9723,
+      "step": 4428
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.775543640794975e-05,
+      "loss": 16.1123,
+      "step": 4429
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.77541051023154e-05,
+      "loss": 18.636,
+      "step": 4430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.775277341105389e-05,
+      "loss": 15.855,
+      "step": 4431
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.775144133417597e-05,
+      "loss": 15.8735,
+      "step": 4432
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.775010887169243e-05,
+      "loss": 14.3236,
+      "step": 4433
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.774877602361398e-05,
+      "loss": 14.6262,
+      "step": 4434
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.774744278995143e-05,
+      "loss": 17.1012,
+      "step": 4435
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.77461091707155e-05,
+      "loss": 14.8295,
+      "step": 4436
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.7744775165917e-05,
+      "loss": 18.5448,
+      "step": 4437
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.77434407755667e-05,
+      "loss": 15.4235,
+      "step": 4438
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.774210599967533e-05,
+      "loss": 15.4727,
+      "step": 4439
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.774077083825372e-05,
+      "loss": 18.4131,
+      "step": 4440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.773943529131265e-05,
+      "loss": 16.9956,
+      "step": 4441
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.773809935886287e-05,
+      "loss": 17.0714,
+      "step": 4442
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.773676304091518e-05,
+      "loss": 17.5405,
+      "step": 4443
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.773542633748039e-05,
+      "loss": 15.8179,
+      "step": 4444
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.773408924856927e-05,
+      "loss": 16.1719,
+      "step": 4445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.773275177419263e-05,
+      "loss": 17.9092,
+      "step": 4446
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.773141391436127e-05,
+      "loss": 17.0079,
+      "step": 4447
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.773007566908599e-05,
+      "loss": 15.9306,
+      "step": 4448
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.77287370383776e-05,
+      "loss": 16.9897,
+      "step": 4449
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.772739802224692e-05,
+      "loss": 17.1032,
+      "step": 4450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.772605862070475e-05,
+      "loss": 15.1951,
+      "step": 4451
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.77247188337619e-05,
+      "loss": 17.5968,
+      "step": 4452
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.77233786614292e-05,
+      "loss": 15.5822,
+      "step": 4453
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.772203810371747e-05,
+      "loss": 16.3436,
+      "step": 4454
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.772069716063753e-05,
+      "loss": 16.6046,
+      "step": 4455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.771935583220022e-05,
+      "loss": 15.9018,
+      "step": 4456
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.771801411841636e-05,
+      "loss": 19.2361,
+      "step": 4457
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.77166720192968e-05,
+      "loss": 16.3204,
+      "step": 4458
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.771532953485235e-05,
+      "loss": 16.5015,
+      "step": 4459
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.771398666509388e-05,
+      "loss": 14.9634,
+      "step": 4460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.771264341003222e-05,
+      "loss": 16.1001,
+      "step": 4461
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.771129976967821e-05,
+      "loss": 14.6926,
+      "step": 4462
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.770995574404273e-05,
+      "loss": 19.0582,
+      "step": 4463
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.77086113331366e-05,
+      "loss": 16.4073,
+      "step": 4464
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.770726653697068e-05,
+      "loss": 17.2061,
+      "step": 4465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.770592135555587e-05,
+      "loss": 14.6962,
+      "step": 4466
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.770457578890298e-05,
+      "loss": 15.3195,
+      "step": 4467
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.770322983702289e-05,
+      "loss": 16.475,
+      "step": 4468
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.770188349992649e-05,
+      "loss": 17.6789,
+      "step": 4469
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.770053677762463e-05,
+      "loss": 15.2872,
+      "step": 4470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.769918967012821e-05,
+      "loss": 15.6268,
+      "step": 4471
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.769784217744808e-05,
+      "loss": 19.0247,
+      "step": 4472
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.769649429959513e-05,
+      "loss": 15.8144,
+      "step": 4473
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.769514603658025e-05,
+      "loss": 15.8399,
+      "step": 4474
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.769379738841433e-05,
+      "loss": 16.9616,
+      "step": 4475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.769244835510826e-05,
+      "loss": 15.2761,
+      "step": 4476
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.769109893667292e-05,
+      "loss": 17.1377,
+      "step": 4477
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.768974913311922e-05,
+      "loss": 15.6311,
+      "step": 4478
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.768839894445805e-05,
+      "loss": 15.912,
+      "step": 4479
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.768704837070032e-05,
+      "loss": 16.0342,
+      "step": 4480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.768569741185695e-05,
+      "loss": 18.6444,
+      "step": 4481
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.768434606793884e-05,
+      "loss": 16.0601,
+      "step": 4482
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.76829943389569e-05,
+      "loss": 17.2586,
+      "step": 4483
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.768164222492202e-05,
+      "loss": 15.9878,
+      "step": 4484
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.768028972584515e-05,
+      "loss": 16.7668,
+      "step": 4485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.767893684173721e-05,
+      "loss": 15.7215,
+      "step": 4486
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.767758357260912e-05,
+      "loss": 17.8158,
+      "step": 4487
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.767622991847181e-05,
+      "loss": 17.4246,
+      "step": 4488
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.76748758793362e-05,
+      "loss": 15.606,
+      "step": 4489
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.767352145521322e-05,
+      "loss": 16.698,
+      "step": 4490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.767216664611386e-05,
+      "loss": 16.9917,
+      "step": 4491
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.767081145204896e-05,
+      "loss": 18.4934,
+      "step": 4492
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.766945587302958e-05,
+      "loss": 15.6712,
+      "step": 4493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.766809990906658e-05,
+      "loss": 14.6881,
+      "step": 4494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.766674356017096e-05,
+      "loss": 19.2615,
+      "step": 4495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.766538682635363e-05,
+      "loss": 16.6142,
+      "step": 4496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.766402970762558e-05,
+      "loss": 16.8221,
+      "step": 4497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.766267220399776e-05,
+      "loss": 16.179,
+      "step": 4498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.766131431548114e-05,
+      "loss": 18.6458,
+      "step": 4499
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.765995604208665e-05,
+      "loss": 15.9148,
+      "step": 4500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.76585973838253e-05,
+      "loss": 16.0815,
+      "step": 4501
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.765723834070804e-05,
+      "loss": 15.3307,
+      "step": 4502
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.765587891274586e-05,
+      "loss": 18.4483,
+      "step": 4503
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.765451909994971e-05,
+      "loss": 16.7328,
+      "step": 4504
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.765315890233062e-05,
+      "loss": 15.6602,
+      "step": 4505
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.765179831989953e-05,
+      "loss": 18.5419,
+      "step": 4506
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.765043735266742e-05,
+      "loss": 17.3022,
+      "step": 4507
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.764907600064531e-05,
+      "loss": 19.4236,
+      "step": 4508
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.764771426384419e-05,
+      "loss": 16.2344,
+      "step": 4509
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.764635214227505e-05,
+      "loss": 18.8598,
+      "step": 4510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.764498963594888e-05,
+      "loss": 15.488,
+      "step": 4511
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.76436267448767e-05,
+      "loss": 16.4416,
+      "step": 4512
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.76422634690695e-05,
+      "loss": 17.5636,
+      "step": 4513
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.76408998085383e-05,
+      "loss": 15.4195,
+      "step": 4514
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.763953576329413e-05,
+      "loss": 17.2688,
+      "step": 4515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.763817133334794e-05,
+      "loss": 16.1141,
+      "step": 4516
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.763680651871082e-05,
+      "loss": 16.9756,
+      "step": 4517
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.763544131939374e-05,
+      "loss": 17.6475,
+      "step": 4518
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.763407573540778e-05,
+      "loss": 16.354,
+      "step": 4519
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.763270976676392e-05,
+      "loss": 17.9924,
+      "step": 4520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.76313434134732e-05,
+      "loss": 16.9237,
+      "step": 4521
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.762997667554667e-05,
+      "loss": 17.7284,
+      "step": 4522
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.762860955299534e-05,
+      "loss": 18.9349,
+      "step": 4523
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.762724204583027e-05,
+      "loss": 17.6546,
+      "step": 4524
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.76258741540625e-05,
+      "loss": 17.0206,
+      "step": 4525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.762450587770308e-05,
+      "loss": 16.1656,
+      "step": 4526
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.762313721676304e-05,
+      "loss": 18.2781,
+      "step": 4527
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.762176817125346e-05,
+      "loss": 17.3395,
+      "step": 4528
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.762039874118538e-05,
+      "loss": 15.0473,
+      "step": 4529
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.761902892656984e-05,
+      "loss": 17.7555,
+      "step": 4530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.761765872741795e-05,
+      "loss": 18.4825,
+      "step": 4531
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.761628814374073e-05,
+      "loss": 17.9272,
+      "step": 4532
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.761491717554927e-05,
+      "loss": 16.9412,
+      "step": 4533
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.761354582285462e-05,
+      "loss": 18.5922,
+      "step": 4534
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.761217408566789e-05,
+      "loss": 17.5705,
+      "step": 4535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.761080196400013e-05,
+      "loss": 16.949,
+      "step": 4536
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.760942945786243e-05,
+      "loss": 17.5135,
+      "step": 4537
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.760805656726586e-05,
+      "loss": 16.5083,
+      "step": 4538
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.760668329222151e-05,
+      "loss": 16.1774,
+      "step": 4539
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.760530963274048e-05,
+      "loss": 16.8198,
+      "step": 4540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.760393558883387e-05,
+      "loss": 17.0807,
+      "step": 4541
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.760256116051276e-05,
+      "loss": 15.7716,
+      "step": 4542
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.760118634778825e-05,
+      "loss": 17.3693,
+      "step": 4543
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.759981115067145e-05,
+      "loss": 17.2797,
+      "step": 4544
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.759843556917346e-05,
+      "loss": 16.1851,
+      "step": 4545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.759705960330537e-05,
+      "loss": 17.164,
+      "step": 4546
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.759568325307833e-05,
+      "loss": 15.7139,
+      "step": 4547
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.759430651850344e-05,
+      "loss": 17.7814,
+      "step": 4548
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.759292939959179e-05,
+      "loss": 16.8309,
+      "step": 4549
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.759155189635453e-05,
+      "loss": 16.611,
+      "step": 4550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.759017400880277e-05,
+      "loss": 16.8078,
+      "step": 4551
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.758879573694766e-05,
+      "loss": 16.0443,
+      "step": 4552
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.758741708080029e-05,
+      "loss": 17.1143,
+      "step": 4553
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.758603804037183e-05,
+      "loss": 15.7403,
+      "step": 4554
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.758465861567341e-05,
+      "loss": 16.1478,
+      "step": 4555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.758327880671614e-05,
+      "loss": 18.7264,
+      "step": 4556
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.75818986135112e-05,
+      "loss": 16.6014,
+      "step": 4557
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.758051803606971e-05,
+      "loss": 16.267,
+      "step": 4558
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.757913707440282e-05,
+      "loss": 19.0973,
+      "step": 4559
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.75777557285217e-05,
+      "loss": 16.6536,
+      "step": 4560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.757637399843749e-05,
+      "loss": 18.3229,
+      "step": 4561
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.757499188416135e-05,
+      "loss": 15.2935,
+      "step": 4562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.757360938570445e-05,
+      "loss": 17.0233,
+      "step": 4563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.757222650307794e-05,
+      "loss": 16.6917,
+      "step": 4564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.7570843236293e-05,
+      "loss": 16.1952,
+      "step": 4565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.756945958536079e-05,
+      "loss": 17.3732,
+      "step": 4566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.756807555029249e-05,
+      "loss": 16.7312,
+      "step": 4567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.756669113109926e-05,
+      "loss": 15.6515,
+      "step": 4568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.75653063277923e-05,
+      "loss": 17.1378,
+      "step": 4569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.75639211403828e-05,
+      "loss": 16.7851,
+      "step": 4570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.756253556888191e-05,
+      "loss": 17.0005,
+      "step": 4571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.756114961330086e-05,
+      "loss": 16.4536,
+      "step": 4572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.755976327365079e-05,
+      "loss": 15.3572,
+      "step": 4573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.755837654994295e-05,
+      "loss": 17.3668,
+      "step": 4574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.755698944218853e-05,
+      "loss": 16.0728,
+      "step": 4575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.75556019503987e-05,
+      "loss": 16.3811,
+      "step": 4576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.755421407458469e-05,
+      "loss": 16.2581,
+      "step": 4577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.755282581475769e-05,
+      "loss": 15.801,
+      "step": 4578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.755143717092892e-05,
+      "loss": 15.0664,
+      "step": 4579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.75500481431096e-05,
+      "loss": 16.2072,
+      "step": 4580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.754865873131092e-05,
+      "loss": 17.7312,
+      "step": 4581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.754726893554413e-05,
+      "loss": 14.2764,
+      "step": 4582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.754587875582046e-05,
+      "loss": 17.487,
+      "step": 4583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.754448819215109e-05,
+      "loss": 16.8627,
+      "step": 4584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.75430972445473e-05,
+      "loss": 18.3844,
+      "step": 4585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.754170591302029e-05,
+      "loss": 15.8132,
+      "step": 4586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.75403141975813e-05,
+      "loss": 17.0659,
+      "step": 4587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.753892209824158e-05,
+      "loss": 17.1292,
+      "step": 4588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.753752961501237e-05,
+      "loss": 16.7793,
+      "step": 4589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.753613674790492e-05,
+      "loss": 17.4592,
+      "step": 4590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.753474349693045e-05,
+      "loss": 18.1454,
+      "step": 4591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.753334986210023e-05,
+      "loss": 17.3856,
+      "step": 4592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.753195584342551e-05,
+      "loss": 17.599,
+      "step": 4593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.753056144091757e-05,
+      "loss": 17.0332,
+      "step": 4594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.752916665458763e-05,
+      "loss": 16.5339,
+      "step": 4595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.752777148444698e-05,
+      "loss": 16.1047,
+      "step": 4596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.752637593050689e-05,
+      "loss": 17.75,
+      "step": 4597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.75249799927786e-05,
+      "loss": 16.826,
+      "step": 4598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.752358367127342e-05,
+      "loss": 17.4821,
+      "step": 4599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.75221869660026e-05,
+      "loss": 16.7991,
+      "step": 4600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.752078987697741e-05,
+      "loss": 16.6382,
+      "step": 4601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.751939240420916e-05,
+      "loss": 19.7192,
+      "step": 4602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.751799454770913e-05,
+      "loss": 16.5053,
+      "step": 4603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.75165963074886e-05,
+      "loss": 15.1722,
+      "step": 4604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.751519768355886e-05,
+      "loss": 18.3376,
+      "step": 4605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.75137986759312e-05,
+      "loss": 17.9349,
+      "step": 4606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.751239928461692e-05,
+      "loss": 16.5491,
+      "step": 4607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.751099950962733e-05,
+      "loss": 18.7684,
+      "step": 4608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.750959935097374e-05,
+      "loss": 17.375,
+      "step": 4609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.750819880866743e-05,
+      "loss": 17.2496,
+      "step": 4610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.750679788271973e-05,
+      "loss": 18.0893,
+      "step": 4611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.750539657314196e-05,
+      "loss": 16.4144,
+      "step": 4612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.75039948799454e-05,
+      "loss": 15.7675,
+      "step": 4613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.750259280314139e-05,
+      "loss": 14.8667,
+      "step": 4614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.750119034274127e-05,
+      "loss": 16.2906,
+      "step": 4615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.749978749875635e-05,
+      "loss": 18.4864,
+      "step": 4616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.749838427119794e-05,
+      "loss": 17.2564,
+      "step": 4617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.749698066007741e-05,
+      "loss": 18.2511,
+      "step": 4618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.749557666540606e-05,
+      "loss": 17.9752,
+      "step": 4619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.749417228719524e-05,
+      "loss": 15.5352,
+      "step": 4620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.749276752545631e-05,
+      "loss": 15.029,
+      "step": 4621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.749136238020057e-05,
+      "loss": 17.6889,
+      "step": 4622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.748995685143942e-05,
+      "loss": 14.8364,
+      "step": 4623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.748855093918417e-05,
+      "loss": 16.2478,
+      "step": 4624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.74871446434462e-05,
+      "loss": 14.7712,
+      "step": 4625
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.748573796423683e-05,
+      "loss": 14.4315,
+      "step": 4626
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.748433090156745e-05,
+      "loss": 17.7281,
+      "step": 4627
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.748292345544941e-05,
+      "loss": 16.2565,
+      "step": 4628
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.74815156258941e-05,
+      "loss": 19.616,
+      "step": 4629
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.748010741291284e-05,
+      "loss": 18.7084,
+      "step": 4630
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.747869881651703e-05,
+      "loss": 17.3788,
+      "step": 4631
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.747728983671806e-05,
+      "loss": 15.4768,
+      "step": 4632
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.74758804735273e-05,
+      "loss": 17.9197,
+      "step": 4633
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.747447072695611e-05,
+      "loss": 17.0554,
+      "step": 4634
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.747306059701588e-05,
+      "loss": 17.3577,
+      "step": 4635
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.747165008371802e-05,
+      "loss": 16.0725,
+      "step": 4636
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.74702391870739e-05,
+      "loss": 15.9361,
+      "step": 4637
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.746882790709491e-05,
+      "loss": 15.4309,
+      "step": 4638
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.746741624379248e-05,
+      "loss": 15.5201,
+      "step": 4639
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.746600419717796e-05,
+      "loss": 15.3313,
+      "step": 4640
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.746459176726278e-05,
+      "loss": 17.7294,
+      "step": 4641
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.746317895405834e-05,
+      "loss": 15.6798,
+      "step": 4642
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.746176575757606e-05,
+      "loss": 15.7471,
+      "step": 4643
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.746035217782734e-05,
+      "loss": 15.906,
+      "step": 4644
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.745893821482361e-05,
+      "loss": 16.1315,
+      "step": 4645
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.745752386857627e-05,
+      "loss": 16.4089,
+      "step": 4646
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.745610913909674e-05,
+      "loss": 16.9575,
+      "step": 4647
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.745469402639647e-05,
+      "loss": 15.8999,
+      "step": 4648
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.745327853048685e-05,
+      "loss": 18.1155,
+      "step": 4649
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.745186265137934e-05,
+      "loss": 16.709,
+      "step": 4650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.745044638908535e-05,
+      "loss": 17.3319,
+      "step": 4651
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.744902974361634e-05,
+      "loss": 17.3143,
+      "step": 4652
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.744761271498375e-05,
+      "loss": 17.8962,
+      "step": 4653
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.744619530319899e-05,
+      "loss": 15.7553,
+      "step": 4654
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.744477750827354e-05,
+      "loss": 17.2717,
+      "step": 4655
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.744335933021885e-05,
+      "loss": 16.732,
+      "step": 4656
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.744194076904636e-05,
+      "loss": 17.3747,
+      "step": 4657
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.744052182476751e-05,
+      "loss": 15.0254,
+      "step": 4658
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.743910249739379e-05,
+      "loss": 16.8715,
+      "step": 4659
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.743768278693663e-05,
+      "loss": 15.2235,
+      "step": 4660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.743626269340751e-05,
+      "loss": 18.5015,
+      "step": 4661
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.743484221681791e-05,
+      "loss": 16.2314,
+      "step": 4662
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.743342135717928e-05,
+      "loss": 16.3764,
+      "step": 4663
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.74320001145031e-05,
+      "loss": 16.9112,
+      "step": 4664
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.743057848880086e-05,
+      "loss": 16.5835,
+      "step": 4665
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.742915648008402e-05,
+      "loss": 16.8873,
+      "step": 4666
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.742773408836407e-05,
+      "loss": 15.9747,
+      "step": 4667
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.742631131365252e-05,
+      "loss": 16.1943,
+      "step": 4668
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.742488815596081e-05,
+      "loss": 17.4127,
+      "step": 4669
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.742346461530048e-05,
+      "loss": 17.7975,
+      "step": 4670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.742204069168299e-05,
+      "loss": 17.6591,
+      "step": 4671
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.742061638511985e-05,
+      "loss": 18.6389,
+      "step": 4672
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.741919169562259e-05,
+      "loss": 16.7355,
+      "step": 4673
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.741776662320268e-05,
+      "loss": 15.8111,
+      "step": 4674
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.741634116787164e-05,
+      "loss": 17.111,
+      "step": 4675
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.741491532964097e-05,
+      "loss": 16.8939,
+      "step": 4676
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.74134891085222e-05,
+      "loss": 16.3023,
+      "step": 4677
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.741206250452683e-05,
+      "loss": 16.8802,
+      "step": 4678
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.741063551766641e-05,
+      "loss": 15.8625,
+      "step": 4679
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.740920814795242e-05,
+      "loss": 15.9908,
+      "step": 4680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.740778039539641e-05,
+      "loss": 16.4639,
+      "step": 4681
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.740635226000993e-05,
+      "loss": 16.0656,
+      "step": 4682
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.74049237418045e-05,
+      "loss": 17.4752,
+      "step": 4683
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.740349484079164e-05,
+      "loss": 16.3498,
+      "step": 4684
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.740206555698288e-05,
+      "loss": 19.0659,
+      "step": 4685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.740063589038978e-05,
+      "loss": 16.7447,
+      "step": 4686
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.73992058410239e-05,
+      "loss": 15.4399,
+      "step": 4687
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.739777540889676e-05,
+      "loss": 17.5545,
+      "step": 4688
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.739634459401992e-05,
+      "loss": 15.8042,
+      "step": 4689
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.739491339640493e-05,
+      "loss": 17.8783,
+      "step": 4690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.739348181606339e-05,
+      "loss": 17.9426,
+      "step": 4691
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.739204985300679e-05,
+      "loss": 17.532,
+      "step": 4692
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.739061750724674e-05,
+      "loss": 15.916,
+      "step": 4693
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.738918477879479e-05,
+      "loss": 17.0868,
+      "step": 4694
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.738775166766251e-05,
+      "loss": 16.6567,
+      "step": 4695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.73863181738615e-05,
+      "loss": 15.8587,
+      "step": 4696
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.73848842974033e-05,
+      "loss": 15.1287,
+      "step": 4697
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.73834500382995e-05,
+      "loss": 17.4636,
+      "step": 4698
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.738201539656168e-05,
+      "loss": 15.9118,
+      "step": 4699
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.738058037220143e-05,
+      "loss": 18.7912,
+      "step": 4700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.737914496523035e-05,
+      "loss": 15.6606,
+      "step": 4701
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.737770917566001e-05,
+      "loss": 16.0487,
+      "step": 4702
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.7376273003502e-05,
+      "loss": 16.2765,
+      "step": 4703
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.737483644876795e-05,
+      "loss": 16.3477,
+      "step": 4704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.737339951146943e-05,
+      "loss": 14.0738,
+      "step": 4705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.737196219161806e-05,
+      "loss": 17.7732,
+      "step": 4706
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.737052448922543e-05,
+      "loss": 16.7438,
+      "step": 4707
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.736908640430318e-05,
+      "loss": 17.7253,
+      "step": 4708
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.73676479368629e-05,
+      "loss": 15.789,
+      "step": 4709
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.73662090869162e-05,
+      "loss": 16.6177,
+      "step": 4710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.736476985447471e-05,
+      "loss": 16.0587,
+      "step": 4711
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.736333023955006e-05,
+      "loss": 17.111,
+      "step": 4712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.736189024215386e-05,
+      "loss": 17.4901,
+      "step": 4713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.736044986229774e-05,
+      "loss": 16.9248,
+      "step": 4714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.735900909999334e-05,
+      "loss": 15.3591,
+      "step": 4715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.735756795525231e-05,
+      "loss": 15.7612,
+      "step": 4716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.735612642808624e-05,
+      "loss": 16.744,
+      "step": 4717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.735468451850681e-05,
+      "loss": 15.3612,
+      "step": 4718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.735324222652566e-05,
+      "loss": 18.3157,
+      "step": 4719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.735179955215442e-05,
+      "loss": 16.5339,
+      "step": 4720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.735035649540476e-05,
+      "loss": 17.2479,
+      "step": 4721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.734891305628832e-05,
+      "loss": 16.6848,
+      "step": 4722
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.734746923481675e-05,
+      "loss": 17.0836,
+      "step": 4723
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.734602503100173e-05,
+      "loss": 15.8323,
+      "step": 4724
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.734458044485491e-05,
+      "loss": 16.4175,
+      "step": 4725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.734313547638795e-05,
+      "loss": 16.1199,
+      "step": 4726
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.734169012561252e-05,
+      "loss": 15.6231,
+      "step": 4727
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.73402443925403e-05,
+      "loss": 16.5408,
+      "step": 4728
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.733879827718297e-05,
+      "loss": 14.6643,
+      "step": 4729
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.733735177955219e-05,
+      "loss": 16.4848,
+      "step": 4730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.733590489965966e-05,
+      "loss": 16.3221,
+      "step": 4731
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.733445763751705e-05,
+      "loss": 20.5499,
+      "step": 4732
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.733300999313604e-05,
+      "loss": 17.585,
+      "step": 4733
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.733156196652835e-05,
+      "loss": 18.0009,
+      "step": 4734
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.733011355770563e-05,
+      "loss": 19.7994,
+      "step": 4735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.732866476667961e-05,
+      "loss": 13.897,
+      "step": 4736
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.732721559346199e-05,
+      "loss": 17.1804,
+      "step": 4737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.732576603806445e-05,
+      "loss": 16.648,
+      "step": 4738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.73243161004987e-05,
+      "loss": 16.5618,
+      "step": 4739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.732286578077647e-05,
+      "loss": 17.193,
+      "step": 4740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.732141507890946e-05,
+      "loss": 19.4835,
+      "step": 4741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.731996399490938e-05,
+      "loss": 18.2356,
+      "step": 4742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.731851252878794e-05,
+      "loss": 16.3053,
+      "step": 4743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.731706068055688e-05,
+      "loss": 16.9906,
+      "step": 4744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.731560845022792e-05,
+      "loss": 14.9334,
+      "step": 4745
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.731415583781278e-05,
+      "loss": 16.2222,
+      "step": 4746
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.73127028433232e-05,
+      "loss": 17.575,
+      "step": 4747
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.73112494667709e-05,
+      "loss": 16.6136,
+      "step": 4748
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.730979570816761e-05,
+      "loss": 17.1979,
+      "step": 4749
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.73083415675251e-05,
+      "loss": 16.157,
+      "step": 4750
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.730688704485508e-05,
+      "loss": 18.5264,
+      "step": 4751
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.730543214016932e-05,
+      "loss": 17.5791,
+      "step": 4752
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.730397685347956e-05,
+      "loss": 16.7018,
+      "step": 4753
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.730252118479755e-05,
+      "loss": 17.1455,
+      "step": 4754
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.730106513413503e-05,
+      "loss": 15.1062,
+      "step": 4755
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.729960870150381e-05,
+      "loss": 16.8641,
+      "step": 4756
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.729815188691559e-05,
+      "loss": 15.9837,
+      "step": 4757
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.729669469038216e-05,
+      "loss": 19.31,
+      "step": 4758
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.729523711191529e-05,
+      "loss": 15.1921,
+      "step": 4759
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.729377915152675e-05,
+      "loss": 16.2788,
+      "step": 4760
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.729232080922829e-05,
+      "loss": 15.5851,
+      "step": 4761
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.729086208503174e-05,
+      "loss": 18.3372,
+      "step": 4762
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.728940297894883e-05,
+      "loss": 15.4431,
+      "step": 4763
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.728794349099135e-05,
+      "loss": 16.5657,
+      "step": 4764
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.728648362117111e-05,
+      "loss": 16.3423,
+      "step": 4765
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.728502336949987e-05,
+      "loss": 14.539,
+      "step": 4766
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.728356273598945e-05,
+      "loss": 16.5108,
+      "step": 4767
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.728210172065162e-05,
+      "loss": 17.2273,
+      "step": 4768
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.72806403234982e-05,
+      "loss": 17.0094,
+      "step": 4769
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.727917854454096e-05,
+      "loss": 14.3275,
+      "step": 4770
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.727771638379175e-05,
+      "loss": 18.6596,
+      "step": 4771
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.727625384126234e-05,
+      "loss": 15.6178,
+      "step": 4772
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.727479091696455e-05,
+      "loss": 15.68,
+      "step": 4773
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.727332761091018e-05,
+      "loss": 16.882,
+      "step": 4774
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.727186392311107e-05,
+      "loss": 16.3185,
+      "step": 4775
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.727039985357905e-05,
+      "loss": 17.3508,
+      "step": 4776
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.726893540232591e-05,
+      "loss": 14.5801,
+      "step": 4777
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.72674705693635e-05,
+      "loss": 15.8677,
+      "step": 4778
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.726600535470363e-05,
+      "loss": 16.686,
+      "step": 4779
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.726453975835814e-05,
+      "loss": 15.6174,
+      "step": 4780
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.726307378033886e-05,
+      "loss": 16.3716,
+      "step": 4781
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.726160742065764e-05,
+      "loss": 17.4058,
+      "step": 4782
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.726014067932632e-05,
+      "loss": 15.7729,
+      "step": 4783
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.725867355635674e-05,
+      "loss": 17.3922,
+      "step": 4784
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.725720605176074e-05,
+      "loss": 16.3418,
+      "step": 4785
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.725573816555019e-05,
+      "loss": 19.347,
+      "step": 4786
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.725426989773693e-05,
+      "loss": 16.5284,
+      "step": 4787
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.72528012483328e-05,
+      "loss": 15.3427,
+      "step": 4788
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.72513322173497e-05,
+      "loss": 17.4058,
+      "step": 4789
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.724986280479947e-05,
+      "loss": 20.1535,
+      "step": 4790
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.724839301069395e-05,
+      "loss": 15.8678,
+      "step": 4791
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.724692283504506e-05,
+      "loss": 16.2643,
+      "step": 4792
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.724545227786464e-05,
+      "loss": 17.3485,
+      "step": 4793
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.724398133916457e-05,
+      "loss": 18.3312,
+      "step": 4794
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.724251001895673e-05,
+      "loss": 18.7492,
+      "step": 4795
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.724103831725302e-05,
+      "loss": 18.4491,
+      "step": 4796
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.723956623406529e-05,
+      "loss": 17.8434,
+      "step": 4797
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.723809376940544e-05,
+      "loss": 16.193,
+      "step": 4798
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.723662092328537e-05,
+      "loss": 15.8195,
+      "step": 4799
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.723514769571697e-05,
+      "loss": 18.3078,
+      "step": 4800
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.723367408671213e-05,
+      "loss": 17.748,
+      "step": 4801
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.723220009628277e-05,
+      "loss": 16.3808,
+      "step": 4802
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.723072572444076e-05,
+      "loss": 18.4354,
+      "step": 4803
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.722925097119803e-05,
+      "loss": 15.2706,
+      "step": 4804
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.722777583656648e-05,
+      "loss": 18.177,
+      "step": 4805
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.722630032055803e-05,
+      "loss": 18.1354,
+      "step": 4806
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.722482442318459e-05,
+      "loss": 18.544,
+      "step": 4807
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.722334814445809e-05,
+      "loss": 16.3591,
+      "step": 4808
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.722187148439041e-05,
+      "loss": 16.8758,
+      "step": 4809
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.722039444299351e-05,
+      "loss": 17.5873,
+      "step": 4810
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.721891702027932e-05,
+      "loss": 16.5626,
+      "step": 4811
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.721743921625976e-05,
+      "loss": 17.3432,
+      "step": 4812
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.721596103094678e-05,
+      "loss": 16.8219,
+      "step": 4813
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.721448246435229e-05,
+      "loss": 15.6033,
+      "step": 4814
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.721300351648823e-05,
+      "loss": 17.3712,
+      "step": 4815
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.721152418736657e-05,
+      "loss": 17.0798,
+      "step": 4816
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.721004447699924e-05,
+      "loss": 17.593,
+      "step": 4817
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.720856438539818e-05,
+      "loss": 15.6713,
+      "step": 4818
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.720708391257537e-05,
+      "loss": 17.3139,
+      "step": 4819
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.720560305854273e-05,
+      "loss": 16.3853,
+      "step": 4820
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.720412182331225e-05,
+      "loss": 15.7943,
+      "step": 4821
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.720264020689587e-05,
+      "loss": 17.0259,
+      "step": 4822
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.720115820930557e-05,
+      "loss": 18.1896,
+      "step": 4823
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.719967583055329e-05,
+      "loss": 17.1867,
+      "step": 4824
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.719819307065103e-05,
+      "loss": 15.8861,
+      "step": 4825
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.719670992961077e-05,
+      "loss": 16.8813,
+      "step": 4826
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.719522640744445e-05,
+      "loss": 15.967,
+      "step": 4827
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.719374250416406e-05,
+      "loss": 16.8943,
+      "step": 4828
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.719225821978162e-05,
+      "loss": 15.0313,
+      "step": 4829
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.719077355430908e-05,
+      "loss": 18.1006,
+      "step": 4830
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.718928850775844e-05,
+      "loss": 15.9686,
+      "step": 4831
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.718780308014167e-05,
+      "loss": 16.7242,
+      "step": 4832
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.71863172714708e-05,
+      "loss": 17.4025,
+      "step": 4833
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.718483108175783e-05,
+      "loss": 17.4524,
+      "step": 4834
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.718334451101472e-05,
+      "loss": 16.7667,
+      "step": 4835
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.718185755925351e-05,
+      "loss": 16.8084,
+      "step": 4836
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.718037022648619e-05,
+      "loss": 17.6991,
+      "step": 4837
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.717888251272477e-05,
+      "loss": 18.2357,
+      "step": 4838
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.717739441798129e-05,
+      "loss": 16.4395,
+      "step": 4839
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.717590594226774e-05,
+      "loss": 18.1805,
+      "step": 4840
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.717441708559616e-05,
+      "loss": 16.8736,
+      "step": 4841
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.717292784797854e-05,
+      "loss": 16.8628,
+      "step": 4842
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.717143822942693e-05,
+      "loss": 17.5062,
+      "step": 4843
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.716994822995338e-05,
+      "loss": 17.8271,
+      "step": 4844
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.716845784956988e-05,
+      "loss": 16.4875,
+      "step": 4845
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.716696708828848e-05,
+      "loss": 17.4161,
+      "step": 4846
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.716547594612121e-05,
+      "loss": 16.4858,
+      "step": 4847
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.716398442308014e-05,
+      "loss": 17.9266,
+      "step": 4848
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.71624925191773e-05,
+      "loss": 16.4885,
+      "step": 4849
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.716100023442474e-05,
+      "loss": 17.334,
+      "step": 4850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.715950756883449e-05,
+      "loss": 17.1924,
+      "step": 4851
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.715801452241863e-05,
+      "loss": 16.3264,
+      "step": 4852
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.71565210951892e-05,
+      "loss": 15.8551,
+      "step": 4853
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.715502728715826e-05,
+      "loss": 17.5882,
+      "step": 4854
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.71535330983379e-05,
+      "loss": 16.4723,
+      "step": 4855
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.715203852874016e-05,
+      "loss": 18.8399,
+      "step": 4856
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.71505435783771e-05,
+      "loss": 17.851,
+      "step": 4857
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.71490482472608e-05,
+      "loss": 16.8444,
+      "step": 4858
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.714755253540335e-05,
+      "loss": 18.5307,
+      "step": 4859
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.714605644281682e-05,
+      "loss": 16.6748,
+      "step": 4860
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.714455996951328e-05,
+      "loss": 17.786,
+      "step": 4861
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.714306311550485e-05,
+      "loss": 15.4194,
+      "step": 4862
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.714156588080357e-05,
+      "loss": 17.1084,
+      "step": 4863
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.714006826542156e-05,
+      "loss": 17.0124,
+      "step": 4864
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.713857026937091e-05,
+      "loss": 16.2797,
+      "step": 4865
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.713707189266371e-05,
+      "loss": 17.9946,
+      "step": 4866
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.713557313531208e-05,
+      "loss": 15.341,
+      "step": 4867
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.713407399732808e-05,
+      "loss": 17.5413,
+      "step": 4868
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.713257447872385e-05,
+      "loss": 16.5449,
+      "step": 4869
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.71310745795115e-05,
+      "loss": 16.7794,
+      "step": 4870
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.712957429970311e-05,
+      "loss": 14.7714,
+      "step": 4871
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.712807363931084e-05,
+      "loss": 18.2261,
+      "step": 4872
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.71265725983468e-05,
+      "loss": 16.2233,
+      "step": 4873
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.712507117682305e-05,
+      "loss": 17.8893,
+      "step": 4874
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.71235693747518e-05,
+      "loss": 16.4172,
+      "step": 4875
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.712206719214514e-05,
+      "loss": 14.8431,
+      "step": 4876
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.712056462901519e-05,
+      "loss": 16.0145,
+      "step": 4877
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.71190616853741e-05,
+      "loss": 18.3903,
+      "step": 4878
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.7117558361234e-05,
+      "loss": 16.116,
+      "step": 4879
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.711605465660701e-05,
+      "loss": 16.2366,
+      "step": 4880
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.711455057150532e-05,
+      "loss": 15.6345,
+      "step": 4881
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.711304610594104e-05,
+      "loss": 17.2567,
+      "step": 4882
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.711154125992632e-05,
+      "loss": 18.9204,
+      "step": 4883
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.711003603347332e-05,
+      "loss": 17.9417,
+      "step": 4884
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.71085304265942e-05,
+      "loss": 17.4664,
+      "step": 4885
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.71070244393011e-05,
+      "loss": 17.4231,
+      "step": 4886
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.710551807160622e-05,
+      "loss": 17.3304,
+      "step": 4887
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.71040113235217e-05,
+      "loss": 17.1128,
+      "step": 4888
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.710250419505968e-05,
+      "loss": 18.6715,
+      "step": 4889
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.710099668623237e-05,
+      "loss": 16.7637,
+      "step": 4890
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.709948879705193e-05,
+      "loss": 17.327,
+      "step": 4891
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.709798052753054e-05,
+      "loss": 16.2593,
+      "step": 4892
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.709647187768036e-05,
+      "loss": 16.9912,
+      "step": 4893
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.709496284751361e-05,
+      "loss": 15.8125,
+      "step": 4894
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.709345343704244e-05,
+      "loss": 19.0911,
+      "step": 4895
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.709194364627907e-05,
+      "loss": 15.5399,
+      "step": 4896
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.709043347523567e-05,
+      "loss": 15.2368,
+      "step": 4897
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.708892292392443e-05,
+      "loss": 15.8532,
+      "step": 4898
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.708741199235758e-05,
+      "loss": 18.199,
+      "step": 4899
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.708590068054728e-05,
+      "loss": 18.4062,
+      "step": 4900
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.708438898850576e-05,
+      "loss": 16.8697,
+      "step": 4901
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.708287691624523e-05,
+      "loss": 18.6347,
+      "step": 4902
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.708136446377787e-05,
+      "loss": 16.9908,
+      "step": 4903
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.707985163111594e-05,
+      "loss": 15.6642,
+      "step": 4904
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.707833841827162e-05,
+      "loss": 18.4945,
+      "step": 4905
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.707682482525714e-05,
+      "loss": 16.5244,
+      "step": 4906
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.707531085208472e-05,
+      "loss": 15.4072,
+      "step": 4907
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.70737964987666e-05,
+      "loss": 15.5215,
+      "step": 4908
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.7072281765315e-05,
+      "loss": 17.5301,
+      "step": 4909
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.707076665174215e-05,
+      "loss": 16.817,
+      "step": 4910
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.70692511580603e-05,
+      "loss": 17.3866,
+      "step": 4911
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.706773528428165e-05,
+      "loss": 18.3288,
+      "step": 4912
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.706621903041847e-05,
+      "loss": 15.4732,
+      "step": 4913
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.7064702396483e-05,
+      "loss": 15.7213,
+      "step": 4914
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.70631853824875e-05,
+      "loss": 19.1198,
+      "step": 4915
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.706166798844418e-05,
+      "loss": 16.931,
+      "step": 4916
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.706015021436534e-05,
+      "loss": 16.9232,
+      "step": 4917
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.705863206026321e-05,
+      "loss": 15.0347,
+      "step": 4918
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.705711352615006e-05,
+      "loss": 15.5223,
+      "step": 4919
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.705559461203815e-05,
+      "loss": 16.8514,
+      "step": 4920
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.705407531793973e-05,
+      "loss": 15.1462,
+      "step": 4921
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.705255564386709e-05,
+      "loss": 17.2973,
+      "step": 4922
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.70510355898325e-05,
+      "loss": 17.062,
+      "step": 4923
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.704951515584822e-05,
+      "loss": 15.0998,
+      "step": 4924
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.704799434192655e-05,
+      "loss": 16.4541,
+      "step": 4925
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.704647314807975e-05,
+      "loss": 16.7762,
+      "step": 4926
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.704495157432011e-05,
+      "loss": 17.1419,
+      "step": 4927
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.704342962065993e-05,
+      "loss": 17.3317,
+      "step": 4928
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.704190728711148e-05,
+      "loss": 17.3764,
+      "step": 4929
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.704038457368706e-05,
+      "loss": 18.9979,
+      "step": 4930
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.703886148039898e-05,
+      "loss": 16.912,
+      "step": 4931
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.703733800725951e-05,
+      "loss": 15.4821,
+      "step": 4932
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.703581415428099e-05,
+      "loss": 17.9615,
+      "step": 4933
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.70342899214757e-05,
+      "loss": 16.4233,
+      "step": 4934
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.703276530885596e-05,
+      "loss": 16.3562,
+      "step": 4935
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.703124031643407e-05,
+      "loss": 18.2506,
+      "step": 4936
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.702971494422235e-05,
+      "loss": 16.1752,
+      "step": 4937
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.702818919223314e-05,
+      "loss": 17.2112,
+      "step": 4938
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.702666306047872e-05,
+      "loss": 15.4494,
+      "step": 4939
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.702513654897143e-05,
+      "loss": 16.1234,
+      "step": 4940
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.70236096577236e-05,
+      "loss": 15.5594,
+      "step": 4941
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.702208238674756e-05,
+      "loss": 16.2715,
+      "step": 4942
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.702055473605566e-05,
+      "loss": 15.244,
+      "step": 4943
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.701902670566023e-05,
+      "loss": 18.272,
+      "step": 4944
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.701749829557357e-05,
+      "loss": 16.4068,
+      "step": 4945
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.701596950580806e-05,
+      "loss": 19.0937,
+      "step": 4946
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.701444033637606e-05,
+      "loss": 15.2439,
+      "step": 4947
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.701291078728987e-05,
+      "loss": 15.4854,
+      "step": 4948
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.701138085856187e-05,
+      "loss": 16.7796,
+      "step": 4949
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.700985055020441e-05,
+      "loss": 17.7686,
+      "step": 4950
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.700831986222987e-05,
+      "loss": 15.7737,
+      "step": 4951
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.700678879465057e-05,
+      "loss": 17.1753,
+      "step": 4952
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.700525734747891e-05,
+      "loss": 15.1781,
+      "step": 4953
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.700372552072723e-05,
+      "loss": 14.7868,
+      "step": 4954
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.700219331440792e-05,
+      "loss": 16.7286,
+      "step": 4955
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.700066072853333e-05,
+      "loss": 16.4698,
+      "step": 4956
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.699912776311587e-05,
+      "loss": 16.754,
+      "step": 4957
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.699759441816787e-05,
+      "loss": 17.267,
+      "step": 4958
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.699606069370175e-05,
+      "loss": 16.1387,
+      "step": 4959
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.69945265897299e-05,
+      "loss": 16.2425,
+      "step": 4960
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.699299210626469e-05,
+      "loss": 16.9236,
+      "step": 4961
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.69914572433185e-05,
+      "loss": 16.8378,
+      "step": 4962
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.698992200090377e-05,
+      "loss": 17.7488,
+      "step": 4963
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.698838637903284e-05,
+      "loss": 16.3904,
+      "step": 4964
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.698685037771815e-05,
+      "loss": 16.6202,
+      "step": 4965
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.698531399697211e-05,
+      "loss": 15.7366,
+      "step": 4966
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.698377723680708e-05,
+      "loss": 16.7112,
+      "step": 4967
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.698224009723551e-05,
+      "loss": 18.6894,
+      "step": 4968
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.69807025782698e-05,
+      "loss": 15.7066,
+      "step": 4969
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.697916467992236e-05,
+      "loss": 15.2017,
+      "step": 4970
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.697762640220564e-05,
+      "loss": 17.1533,
+      "step": 4971
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.697608774513202e-05,
+      "loss": 14.3435,
+      "step": 4972
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.697454870871395e-05,
+      "loss": 16.8678,
+      "step": 4973
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.697300929296384e-05,
+      "loss": 16.5395,
+      "step": 4974
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.697146949789413e-05,
+      "loss": 15.8506,
+      "step": 4975
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.696992932351727e-05,
+      "loss": 15.268,
+      "step": 4976
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.696838876984569e-05,
+      "loss": 16.9205,
+      "step": 4977
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.69668478368918e-05,
+      "loss": 17.5617,
+      "step": 4978
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.696530652466809e-05,
+      "loss": 18.2122,
+      "step": 4979
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.696376483318699e-05,
+      "loss": 17.4455,
+      "step": 4980
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.696222276246091e-05,
+      "loss": 16.4364,
+      "step": 4981
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.696068031250236e-05,
+      "loss": 17.2114,
+      "step": 4982
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.695913748332378e-05,
+      "loss": 15.8664,
+      "step": 4983
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.695759427493761e-05,
+      "loss": 20.1204,
+      "step": 4984
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.695605068735633e-05,
+      "loss": 16.3562,
+      "step": 4985
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.695450672059239e-05,
+      "loss": 16.5128,
+      "step": 4986
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.695296237465827e-05,
+      "loss": 17.242,
+      "step": 4987
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.695141764956643e-05,
+      "loss": 16.5246,
+      "step": 4988
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.694987254532936e-05,
+      "loss": 16.2359,
+      "step": 4989
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.694832706195951e-05,
+      "loss": 14.5837,
+      "step": 4990
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.69467811994694e-05,
+      "loss": 18.2463,
+      "step": 4991
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.694523495787149e-05,
+      "loss": 16.8985,
+      "step": 4992
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.694368833717825e-05,
+      "loss": 18.6629,
+      "step": 4993
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.69421413374022e-05,
+      "loss": 15.0721,
+      "step": 4994
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.694059395855582e-05,
+      "loss": 18.3892,
+      "step": 4995
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.69390462006516e-05,
+      "loss": 16.8213,
+      "step": 4996
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.693749806370205e-05,
+      "loss": 19.0372,
+      "step": 4997
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.693594954771965e-05,
+      "loss": 16.5785,
+      "step": 4998
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.693440065271693e-05,
+      "loss": 16.8216,
+      "step": 4999
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.69328513787064e-05,
+      "loss": 16.332,
+      "step": 5000
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.693130172570055e-05,
+      "loss": 16.3715,
+      "step": 5001
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.69297516937119e-05,
+      "loss": 16.3288,
+      "step": 5002
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.692820128275295e-05,
+      "loss": 17.918,
+      "step": 5003
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.692665049283627e-05,
+      "loss": 19.1924,
+      "step": 5004
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.692509932397434e-05,
+      "loss": 18.1137,
+      "step": 5005
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.692354777617969e-05,
+      "loss": 17.7896,
+      "step": 5006
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.692199584946486e-05,
+      "loss": 17.923,
+      "step": 5007
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.69204435438424e-05,
+      "loss": 15.9199,
+      "step": 5008
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.691889085932481e-05,
+      "loss": 17.3137,
+      "step": 5009
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.691733779592465e-05,
+      "loss": 15.2987,
+      "step": 5010
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.691578435365444e-05,
+      "loss": 14.4563,
+      "step": 5011
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.691423053252674e-05,
+      "loss": 18.2473,
+      "step": 5012
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.691267633255411e-05,
+      "loss": 17.919,
+      "step": 5013
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.691112175374909e-05,
+      "loss": 17.6981,
+      "step": 5014
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.690956679612421e-05,
+      "loss": 15.9574,
+      "step": 5015
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.690801145969206e-05,
+      "loss": 16.104,
+      "step": 5016
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.69064557444652e-05,
+      "loss": 16.142,
+      "step": 5017
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.690489965045615e-05,
+      "loss": 16.3797,
+      "step": 5018
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.690334317767754e-05,
+      "loss": 16.2335,
+      "step": 5019
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.690178632614188e-05,
+      "loss": 16.2939,
+      "step": 5020
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.690022909586179e-05,
+      "loss": 14.4489,
+      "step": 5021
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.689867148684979e-05,
+      "loss": 16.6934,
+      "step": 5022
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.689711349911851e-05,
+      "loss": 14.4778,
+      "step": 5023
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.689555513268052e-05,
+      "loss": 18.259,
+      "step": 5024
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.689399638754836e-05,
+      "loss": 16.7018,
+      "step": 5025
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.68924372637347e-05,
+      "loss": 18.3236,
+      "step": 5026
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.689087776125206e-05,
+      "loss": 15.5085,
+      "step": 5027
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.688931788011305e-05,
+      "loss": 16.3033,
+      "step": 5028
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.688775762033029e-05,
+      "loss": 16.9271,
+      "step": 5029
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.688619698191634e-05,
+      "loss": 18.2934,
+      "step": 5030
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.688463596488384e-05,
+      "loss": 18.238,
+      "step": 5031
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.688307456924538e-05,
+      "loss": 16.376,
+      "step": 5032
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.688151279501357e-05,
+      "loss": 16.6842,
+      "step": 5033
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.687995064220102e-05,
+      "loss": 16.3486,
+      "step": 5034
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.687838811082035e-05,
+      "loss": 18.5317,
+      "step": 5035
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.687682520088417e-05,
+      "loss": 16.2232,
+      "step": 5036
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.68752619124051e-05,
+      "loss": 16.1691,
+      "step": 5037
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.687369824539577e-05,
+      "loss": 16.9916,
+      "step": 5038
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.687213419986882e-05,
+      "loss": 16.6726,
+      "step": 5039
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.687056977583684e-05,
+      "loss": 16.2419,
+      "step": 5040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.68690049733125e-05,
+      "loss": 17.6133,
+      "step": 5041
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.686743979230844e-05,
+      "loss": 16.4277,
+      "step": 5042
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.686587423283727e-05,
+      "loss": 17.9387,
+      "step": 5043
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.686430829491165e-05,
+      "loss": 17.828,
+      "step": 5044
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.686274197854422e-05,
+      "loss": 16.4661,
+      "step": 5045
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.686117528374765e-05,
+      "loss": 16.6444,
+      "step": 5046
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.685960821053455e-05,
+      "loss": 17.2985,
+      "step": 5047
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.68580407589176e-05,
+      "loss": 15.9642,
+      "step": 5048
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.685647292890945e-05,
+      "loss": 17.1045,
+      "step": 5049
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.685490472052276e-05,
+      "loss": 15.1705,
+      "step": 5050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.685333613377021e-05,
+      "loss": 17.4098,
+      "step": 5051
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.685176716866446e-05,
+      "loss": 17.5171,
+      "step": 5052
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.685019782521817e-05,
+      "loss": 15.921,
+      "step": 5053
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.684862810344402e-05,
+      "loss": 15.0018,
+      "step": 5054
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.684705800335467e-05,
+      "loss": 17.166,
+      "step": 5055
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.684548752496284e-05,
+      "loss": 16.4403,
+      "step": 5056
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.684391666828117e-05,
+      "loss": 14.7054,
+      "step": 5057
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.684234543332236e-05,
+      "loss": 15.0975,
+      "step": 5058
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.684077382009909e-05,
+      "loss": 16.0572,
+      "step": 5059
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.683920182862407e-05,
+      "loss": 17.3513,
+      "step": 5060
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.683762945890997e-05,
+      "loss": 17.1391,
+      "step": 5061
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.683605671096952e-05,
+      "loss": 17.4054,
+      "step": 5062
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.683448358481538e-05,
+      "loss": 15.3214,
+      "step": 5063
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.683291008046029e-05,
+      "loss": 17.3719,
+      "step": 5064
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.683133619791693e-05,
+      "loss": 15.8435,
+      "step": 5065
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.682976193719803e-05,
+      "loss": 17.0293,
+      "step": 5066
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.68281872983163e-05,
+      "loss": 16.7653,
+      "step": 5067
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.682661228128444e-05,
+      "loss": 15.9361,
+      "step": 5068
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.682503688611517e-05,
+      "loss": 17.3937,
+      "step": 5069
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.682346111282123e-05,
+      "loss": 20.1374,
+      "step": 5070
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.682188496141533e-05,
+      "loss": 16.2144,
+      "step": 5071
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.682030843191022e-05,
+      "loss": 16.2553,
+      "step": 5072
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.681873152431857e-05,
+      "loss": 16.9733,
+      "step": 5073
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.681715423865319e-05,
+      "loss": 16.4715,
+      "step": 5074
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.681557657492679e-05,
+      "loss": 16.9875,
+      "step": 5075
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.681399853315207e-05,
+      "loss": 17.8182,
+      "step": 5076
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.681242011334184e-05,
+      "loss": 17.1132,
+      "step": 5077
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.68108413155088e-05,
+      "loss": 17.9062,
+      "step": 5078
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.68092621396657e-05,
+      "loss": 18.266,
+      "step": 5079
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.680768258582532e-05,
+      "loss": 16.6901,
+      "step": 5080
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.680610265400039e-05,
+      "loss": 16.6164,
+      "step": 5081
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.68045223442037e-05,
+      "loss": 16.5012,
+      "step": 5082
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.680294165644797e-05,
+      "loss": 17.2221,
+      "step": 5083
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.680136059074598e-05,
+      "loss": 18.0602,
+      "step": 5084
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.679977914711051e-05,
+      "loss": 18.1353,
+      "step": 5085
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.679819732555432e-05,
+      "loss": 17.8901,
+      "step": 5086
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.67966151260902e-05,
+      "loss": 19.7918,
+      "step": 5087
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.67950325487309e-05,
+      "loss": 15.0943,
+      "step": 5088
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.679344959348922e-05,
+      "loss": 17.9313,
+      "step": 5089
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.679186626037791e-05,
+      "loss": 15.1447,
+      "step": 5090
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.67902825494098e-05,
+      "loss": 16.3703,
+      "step": 5091
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.678869846059766e-05,
+      "loss": 14.2243,
+      "step": 5092
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.678711399395428e-05,
+      "loss": 16.9037,
+      "step": 5093
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.678552914949245e-05,
+      "loss": 16.1054,
+      "step": 5094
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.6783943927225e-05,
+      "loss": 16.8933,
+      "step": 5095
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.678235832716468e-05,
+      "loss": 18.9999,
+      "step": 5096
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.678077234932433e-05,
+      "loss": 17.1832,
+      "step": 5097
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.677918599371674e-05,
+      "loss": 16.8354,
+      "step": 5098
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.677759926035473e-05,
+      "loss": 15.7652,
+      "step": 5099
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.677601214925111e-05,
+      "loss": 16.0805,
+      "step": 5100
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.67744246604187e-05,
+      "loss": 16.2215,
+      "step": 5101
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.677283679387032e-05,
+      "loss": 15.5344,
+      "step": 5102
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.677124854961879e-05,
+      "loss": 17.3831,
+      "step": 5103
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.676965992767693e-05,
+      "loss": 18.1225,
+      "step": 5104
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.676807092805757e-05,
+      "loss": 16.4677,
+      "step": 5105
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.676648155077354e-05,
+      "loss": 15.9206,
+      "step": 5106
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.67648917958377e-05,
+      "loss": 16.8453,
+      "step": 5107
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.676330166326285e-05,
+      "loss": 18.2371,
+      "step": 5108
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.676171115306186e-05,
+      "loss": 14.7439,
+      "step": 5109
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.676012026524755e-05,
+      "loss": 16.2546,
+      "step": 5110
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.675852899983277e-05,
+      "loss": 15.9306,
+      "step": 5111
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.67569373568304e-05,
+      "loss": 15.2822,
+      "step": 5112
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.675534533625325e-05,
+      "loss": 16.1353,
+      "step": 5113
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.675375293811421e-05,
+      "loss": 16.2855,
+      "step": 5114
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.675216016242613e-05,
+      "loss": 16.1905,
+      "step": 5115
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.675056700920186e-05,
+      "loss": 16.2273,
+      "step": 5116
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.674897347845428e-05,
+      "loss": 16.3894,
+      "step": 5117
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.674737957019624e-05,
+      "loss": 14.8556,
+      "step": 5118
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.674578528444062e-05,
+      "loss": 15.5044,
+      "step": 5119
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.674419062120031e-05,
+      "loss": 17.4436,
+      "step": 5120
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.674259558048817e-05,
+      "loss": 17.6474,
+      "step": 5121
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.67410001623171e-05,
+      "loss": 18.0847,
+      "step": 5122
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.673940436669995e-05,
+      "loss": 16.1005,
+      "step": 5123
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.673780819364962e-05,
+      "loss": 16.0543,
+      "step": 5124
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.673621164317902e-05,
+      "loss": 18.0542,
+      "step": 5125
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.673461471530101e-05,
+      "loss": 16.9107,
+      "step": 5126
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.673301741002852e-05,
+      "loss": 17.7757,
+      "step": 5127
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.673141972737443e-05,
+      "loss": 17.4937,
+      "step": 5128
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.672982166735163e-05,
+      "loss": 17.3764,
+      "step": 5129
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.672822322997305e-05,
+      "loss": 15.8403,
+      "step": 5130
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.672662441525158e-05,
+      "loss": 16.9145,
+      "step": 5131
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.672502522320012e-05,
+      "loss": 16.6236,
+      "step": 5132
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.672342565383163e-05,
+      "loss": 17.1725,
+      "step": 5133
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.672182570715898e-05,
+      "loss": 14.1396,
+      "step": 5134
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.672022538319512e-05,
+      "loss": 15.9101,
+      "step": 5135
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.671862468195294e-05,
+      "loss": 16.0011,
+      "step": 5136
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.671702360344539e-05,
+      "loss": 15.7566,
+      "step": 5137
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.67154221476854e-05,
+      "loss": 17.0822,
+      "step": 5138
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.671382031468591e-05,
+      "loss": 15.4309,
+      "step": 5139
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.671221810445983e-05,
+      "loss": 17.0545,
+      "step": 5140
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.671061551702013e-05,
+      "loss": 17.5266,
+      "step": 5141
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.67090125523797e-05,
+      "loss": 17.2117,
+      "step": 5142
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.670740921055152e-05,
+      "loss": 16.1037,
+      "step": 5143
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.670580549154855e-05,
+      "loss": 18.1145,
+      "step": 5144
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.670420139538372e-05,
+      "loss": 15.8989,
+      "step": 5145
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.670259692207e-05,
+      "loss": 17.9851,
+      "step": 5146
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.670099207162031e-05,
+      "loss": 15.2971,
+      "step": 5147
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.669938684404766e-05,
+      "loss": 17.3798,
+      "step": 5148
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.669778123936497e-05,
+      "loss": 16.2845,
+      "step": 5149
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.669617525758523e-05,
+      "loss": 14.2572,
+      "step": 5150
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.669456889872139e-05,
+      "loss": 16.7634,
+      "step": 5151
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.669296216278644e-05,
+      "loss": 15.7984,
+      "step": 5152
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.669135504979336e-05,
+      "loss": 17.9161,
+      "step": 5153
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.66897475597551e-05,
+      "loss": 16.9862,
+      "step": 5154
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.668813969268466e-05,
+      "loss": 15.9521,
+      "step": 5155
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.668653144859503e-05,
+      "loss": 16.224,
+      "step": 5156
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.668492282749917e-05,
+      "loss": 15.1936,
+      "step": 5157
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.66833138294101e-05,
+      "loss": 17.5134,
+      "step": 5158
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.66817044543408e-05,
+      "loss": 16.1811,
+      "step": 5159
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.668009470230427e-05,
+      "loss": 19.3434,
+      "step": 5160
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.66784845733135e-05,
+      "loss": 16.7246,
+      "step": 5161
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.667687406738151e-05,
+      "loss": 17.1018,
+      "step": 5162
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.667526318452127e-05,
+      "loss": 18.6792,
+      "step": 5163
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.667365192474584e-05,
+      "loss": 15.1119,
+      "step": 5164
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.667204028806818e-05,
+      "loss": 16.4874,
+      "step": 5165
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.667042827450133e-05,
+      "loss": 16.0076,
+      "step": 5166
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.666881588405832e-05,
+      "loss": 17.6208,
+      "step": 5167
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.666720311675215e-05,
+      "loss": 16.4469,
+      "step": 5168
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.666558997259585e-05,
+      "loss": 15.938,
+      "step": 5169
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.666397645160243e-05,
+      "loss": 16.6913,
+      "step": 5170
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.666236255378496e-05,
+      "loss": 16.3976,
+      "step": 5171
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.666074827915644e-05,
+      "loss": 17.1585,
+      "step": 5172
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.665913362772989e-05,
+      "loss": 16.6881,
+      "step": 5173
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.66575185995184e-05,
+      "loss": 17.8657,
+      "step": 5174
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.665590319453497e-05,
+      "loss": 16.3714,
+      "step": 5175
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.665428741279266e-05,
+      "loss": 15.9842,
+      "step": 5176
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.665267125430452e-05,
+      "loss": 15.0475,
+      "step": 5177
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.66510547190836e-05,
+      "loss": 15.0166,
+      "step": 5178
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.664943780714294e-05,
+      "loss": 16.9108,
+      "step": 5179
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.664782051849563e-05,
+      "loss": 16.6763,
+      "step": 5180
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.66462028531547e-05,
+      "loss": 17.1748,
+      "step": 5181
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.664458481113323e-05,
+      "loss": 15.7332,
+      "step": 5182
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.664296639244426e-05,
+      "loss": 17.142,
+      "step": 5183
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.664134759710088e-05,
+      "loss": 17.5045,
+      "step": 5184
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.663972842511617e-05,
+      "loss": 16.4459,
+      "step": 5185
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.663810887650318e-05,
+      "loss": 17.6458,
+      "step": 5186
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.663648895127502e-05,
+      "loss": 16.9172,
+      "step": 5187
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.663486864944474e-05,
+      "loss": 17.583,
+      "step": 5188
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.663324797102545e-05,
+      "loss": 16.4488,
+      "step": 5189
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.663162691603023e-05,
+      "loss": 16.3948,
+      "step": 5190
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.663000548447214e-05,
+      "loss": 15.8178,
+      "step": 5191
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.662838367636432e-05,
+      "loss": 15.6968,
+      "step": 5192
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.662676149171985e-05,
+      "loss": 14.8848,
+      "step": 5193
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.66251389305518e-05,
+      "loss": 16.9137,
+      "step": 5194
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.662351599287333e-05,
+      "loss": 16.7076,
+      "step": 5195
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.66218926786975e-05,
+      "loss": 16.8708,
+      "step": 5196
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.662026898803744e-05,
+      "loss": 15.9878,
+      "step": 5197
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.661864492090625e-05,
+      "loss": 16.8074,
+      "step": 5198
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.661702047731704e-05,
+      "loss": 16.7929,
+      "step": 5199
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.661539565728295e-05,
+      "loss": 17.7573,
+      "step": 5200
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.661377046081709e-05,
+      "loss": 17.5163,
+      "step": 5201
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.661214488793257e-05,
+      "loss": 16.9637,
+      "step": 5202
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.661051893864254e-05,
+      "loss": 16.2639,
+      "step": 5203
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.660889261296012e-05,
+      "loss": 15.8371,
+      "step": 5204
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.660726591089844e-05,
+      "loss": 18.4899,
+      "step": 5205
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.660563883247062e-05,
+      "loss": 18.7363,
+      "step": 5206
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.660401137768985e-05,
+      "loss": 17.7267,
+      "step": 5207
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.660238354656921e-05,
+      "loss": 17.5224,
+      "step": 5208
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.660075533912189e-05,
+      "loss": 17.6883,
+      "step": 5209
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.659912675536101e-05,
+      "loss": 16.88,
+      "step": 5210
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.659749779529974e-05,
+      "loss": 14.9781,
+      "step": 5211
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.659586845895123e-05,
+      "loss": 18.4551,
+      "step": 5212
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.659423874632865e-05,
+      "loss": 16.9273,
+      "step": 5213
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.659260865744512e-05,
+      "loss": 17.292,
+      "step": 5214
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.659097819231385e-05,
+      "loss": 16.5132,
+      "step": 5215
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.658934735094798e-05,
+      "loss": 17.2138,
+      "step": 5216
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.658771613336068e-05,
+      "loss": 16.8863,
+      "step": 5217
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.658608453956513e-05,
+      "loss": 16.4454,
+      "step": 5218
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.658445256957449e-05,
+      "loss": 15.6596,
+      "step": 5219
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.658282022340197e-05,
+      "loss": 14.9503,
+      "step": 5220
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.658118750106073e-05,
+      "loss": 15.7492,
+      "step": 5221
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.657955440256395e-05,
+      "loss": 16.5412,
+      "step": 5222
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.657792092792484e-05,
+      "loss": 17.3748,
+      "step": 5223
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.657628707715655e-05,
+      "loss": 15.7362,
+      "step": 5224
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.657465285027234e-05,
+      "loss": 16.6214,
+      "step": 5225
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.657301824728534e-05,
+      "loss": 16.0536,
+      "step": 5226
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.657138326820878e-05,
+      "loss": 16.7911,
+      "step": 5227
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.656974791305586e-05,
+      "loss": 15.6067,
+      "step": 5228
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.65681121818398e-05,
+      "loss": 16.8852,
+      "step": 5229
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.656647607457379e-05,
+      "loss": 18.4141,
+      "step": 5230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.656483959127102e-05,
+      "loss": 15.8711,
+      "step": 5231
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.656320273194477e-05,
+      "loss": 14.6176,
+      "step": 5232
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.65615654966082e-05,
+      "loss": 15.83,
+      "step": 5233
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.655992788527455e-05,
+      "loss": 14.5101,
+      "step": 5234
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.655828989795706e-05,
+      "loss": 14.8863,
+      "step": 5235
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.655665153466892e-05,
+      "loss": 17.3395,
+      "step": 5236
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.65550127954234e-05,
+      "loss": 16.6131,
+      "step": 5237
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.655337368023371e-05,
+      "loss": 16.2929,
+      "step": 5238
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.65517341891131e-05,
+      "loss": 13.8744,
+      "step": 5239
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.655009432207478e-05,
+      "loss": 17.0956,
+      "step": 5240
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.654845407913203e-05,
+      "loss": 16.8821,
+      "step": 5241
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.654681346029808e-05,
+      "loss": 18.1992,
+      "step": 5242
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.654517246558617e-05,
+      "loss": 15.6583,
+      "step": 5243
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.654353109500956e-05,
+      "loss": 16.6389,
+      "step": 5244
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.654188934858151e-05,
+      "loss": 16.7074,
+      "step": 5245
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.654024722631527e-05,
+      "loss": 16.8872,
+      "step": 5246
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.65386047282241e-05,
+      "loss": 16.4224,
+      "step": 5247
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.653696185432127e-05,
+      "loss": 14.8079,
+      "step": 5248
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.653531860462005e-05,
+      "loss": 15.5036,
+      "step": 5249
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.653367497913368e-05,
+      "loss": 16.4913,
+      "step": 5250
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.653203097787546e-05,
+      "loss": 16.5349,
+      "step": 5251
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.653038660085867e-05,
+      "loss": 17.0337,
+      "step": 5252
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.652874184809656e-05,
+      "loss": 15.9223,
+      "step": 5253
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.652709671960245e-05,
+      "loss": 17.8506,
+      "step": 5254
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.652545121538961e-05,
+      "loss": 17.1817,
+      "step": 5255
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.652380533547132e-05,
+      "loss": 16.8393,
+      "step": 5256
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.652215907986087e-05,
+      "loss": 17.0536,
+      "step": 5257
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.652051244857154e-05,
+      "loss": 16.5947,
+      "step": 5258
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.651886544161667e-05,
+      "loss": 15.7019,
+      "step": 5259
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.651721805900953e-05,
+      "loss": 16.6998,
+      "step": 5260
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.651557030076341e-05,
+      "loss": 15.2781,
+      "step": 5261
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.651392216689165e-05,
+      "loss": 17.3827,
+      "step": 5262
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.651227365740755e-05,
+      "loss": 18.2511,
+      "step": 5263
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.651062477232442e-05,
+      "loss": 15.8768,
+      "step": 5264
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.650897551165556e-05,
+      "loss": 15.2821,
+      "step": 5265
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.65073258754143e-05,
+      "loss": 17.4611,
+      "step": 5266
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.650567586361396e-05,
+      "loss": 18.388,
+      "step": 5267
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.650402547626786e-05,
+      "loss": 16.0884,
+      "step": 5268
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.650237471338934e-05,
+      "loss": 16.5801,
+      "step": 5269
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.650072357499173e-05,
+      "loss": 15.8575,
+      "step": 5270
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.649907206108835e-05,
+      "loss": 16.1703,
+      "step": 5271
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.649742017169253e-05,
+      "loss": 15.3186,
+      "step": 5272
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.649576790681762e-05,
+      "loss": 15.9027,
+      "step": 5273
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.649411526647697e-05,
+      "loss": 15.3941,
+      "step": 5274
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.649246225068393e-05,
+      "loss": 16.2328,
+      "step": 5275
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.649080885945184e-05,
+      "loss": 16.2056,
+      "step": 5276
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.648915509279404e-05,
+      "loss": 16.9148,
+      "step": 5277
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.64875009507239e-05,
+      "loss": 16.0691,
+      "step": 5278
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.648584643325477e-05,
+      "loss": 16.0251,
+      "step": 5279
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.64841915404e-05,
+      "loss": 16.6892,
+      "step": 5280
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.648253627217298e-05,
+      "loss": 17.5839,
+      "step": 5281
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.648088062858707e-05,
+      "loss": 14.013,
+      "step": 5282
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.647922460965562e-05,
+      "loss": 17.5152,
+      "step": 5283
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.647756821539202e-05,
+      "loss": 17.0935,
+      "step": 5284
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.647591144580965e-05,
+      "loss": 15.0382,
+      "step": 5285
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.647425430092187e-05,
+      "loss": 16.3245,
+      "step": 5286
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.647259678074207e-05,
+      "loss": 15.3076,
+      "step": 5287
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.647093888528365e-05,
+      "loss": 16.1908,
+      "step": 5288
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.646928061455998e-05,
+      "loss": 16.6503,
+      "step": 5289
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.646762196858445e-05,
+      "loss": 17.6397,
+      "step": 5290
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.646596294737046e-05,
+      "loss": 17.8252,
+      "step": 5291
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.646430355093141e-05,
+      "loss": 15.8354,
+      "step": 5292
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.64626437792807e-05,
+      "loss": 14.3165,
+      "step": 5293
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.646098363243173e-05,
+      "loss": 15.0316,
+      "step": 5294
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.645932311039789e-05,
+      "loss": 16.4667,
+      "step": 5295
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.645766221319264e-05,
+      "loss": 16.6032,
+      "step": 5296
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.645600094082932e-05,
+      "loss": 15.304,
+      "step": 5297
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.64543392933214e-05,
+      "loss": 17.9678,
+      "step": 5298
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.645267727068229e-05,
+      "loss": 15.9451,
+      "step": 5299
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.645101487292539e-05,
+      "loss": 14.9895,
+      "step": 5300
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.644935210006414e-05,
+      "loss": 15.8882,
+      "step": 5301
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.644768895211197e-05,
+      "loss": 17.9838,
+      "step": 5302
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.644602542908229e-05,
+      "loss": 15.4627,
+      "step": 5303
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.644436153098856e-05,
+      "loss": 15.7634,
+      "step": 5304
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.64426972578442e-05,
+      "loss": 19.1589,
+      "step": 5305
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.644103260966266e-05,
+      "loss": 16.8408,
+      "step": 5306
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.643936758645738e-05,
+      "loss": 18.737,
+      "step": 5307
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.64377021882418e-05,
+      "loss": 15.394,
+      "step": 5308
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.643603641502937e-05,
+      "loss": 15.7283,
+      "step": 5309
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.643437026683353e-05,
+      "loss": 18.1422,
+      "step": 5310
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.643270374366775e-05,
+      "loss": 15.8173,
+      "step": 5311
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.64310368455455e-05,
+      "loss": 14.3775,
+      "step": 5312
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.642936957248021e-05,
+      "loss": 16.1001,
+      "step": 5313
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.642770192448536e-05,
+      "loss": 16.4739,
+      "step": 5314
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.64260339015744e-05,
+      "loss": 16.4453,
+      "step": 5315
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.642436550376085e-05,
+      "loss": 17.5822,
+      "step": 5316
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.642269673105812e-05,
+      "loss": 17.7647,
+      "step": 5317
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.642102758347973e-05,
+      "loss": 15.5639,
+      "step": 5318
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.641935806103914e-05,
+      "loss": 19.1327,
+      "step": 5319
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.641768816374983e-05,
+      "loss": 16.335,
+      "step": 5320
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.641601789162528e-05,
+      "loss": 17.8791,
+      "step": 5321
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.641434724467899e-05,
+      "loss": 17.5558,
+      "step": 5322
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.641267622292447e-05,
+      "loss": 18.1767,
+      "step": 5323
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.641100482637518e-05,
+      "loss": 17.3284,
+      "step": 5324
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.640933305504461e-05,
+      "loss": 17.5741,
+      "step": 5325
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.640766090894629e-05,
+      "loss": 17.7352,
+      "step": 5326
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.640598838809371e-05,
+      "loss": 16.5727,
+      "step": 5327
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.640431549250039e-05,
+      "loss": 17.181,
+      "step": 5328
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.640264222217982e-05,
+      "loss": 17.8703,
+      "step": 5329
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.640096857714552e-05,
+      "loss": 15.576,
+      "step": 5330
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.6399294557411e-05,
+      "loss": 16.7479,
+      "step": 5331
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.639762016298979e-05,
+      "loss": 17.0571,
+      "step": 5332
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.639594539389539e-05,
+      "loss": 16.2074,
+      "step": 5333
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.639427025014134e-05,
+      "loss": 17.9027,
+      "step": 5334
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.639259473174117e-05,
+      "loss": 18.1573,
+      "step": 5335
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.639091883870841e-05,
+      "loss": 16.3742,
+      "step": 5336
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.638924257105659e-05,
+      "loss": 16.1798,
+      "step": 5337
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.638756592879922e-05,
+      "loss": 17.054,
+      "step": 5338
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.638588891194989e-05,
+      "loss": 18.2058,
+      "step": 5339
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.63842115205221e-05,
+      "loss": 15.5206,
+      "step": 5340
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.638253375452944e-05,
+      "loss": 15.9124,
+      "step": 5341
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.63808556139854e-05,
+      "loss": 18.4143,
+      "step": 5342
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.637917709890357e-05,
+      "loss": 15.6574,
+      "step": 5343
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.63774982092975e-05,
+      "loss": 15.2503,
+      "step": 5344
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.637581894518075e-05,
+      "loss": 16.7233,
+      "step": 5345
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.637413930656687e-05,
+      "loss": 16.2167,
+      "step": 5346
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.63724592934694e-05,
+      "loss": 16.7342,
+      "step": 5347
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.637077890590198e-05,
+      "loss": 18.7576,
+      "step": 5348
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.636909814387811e-05,
+      "loss": 17.04,
+      "step": 5349
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.636741700741138e-05,
+      "loss": 17.4053,
+      "step": 5350
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.636573549651538e-05,
+      "loss": 18.5558,
+      "step": 5351
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.636405361120368e-05,
+      "loss": 17.4681,
+      "step": 5352
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.636237135148987e-05,
+      "loss": 15.6134,
+      "step": 5353
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.636068871738753e-05,
+      "loss": 15.6812,
+      "step": 5354
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.635900570891024e-05,
+      "loss": 17.4512,
+      "step": 5355
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.635732232607158e-05,
+      "loss": 15.8459,
+      "step": 5356
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.635563856888517e-05,
+      "loss": 19.7267,
+      "step": 5357
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.63539544373646e-05,
+      "loss": 15.6706,
+      "step": 5358
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.635226993152345e-05,
+      "loss": 16.9252,
+      "step": 5359
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.635058505137536e-05,
+      "loss": 16.4867,
+      "step": 5360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.63488997969339e-05,
+      "loss": 17.0127,
+      "step": 5361
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.63472141682127e-05,
+      "loss": 15.9089,
+      "step": 5362
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.634552816522537e-05,
+      "loss": 15.0319,
+      "step": 5363
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.63438417879855e-05,
+      "loss": 17.078,
+      "step": 5364
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.634215503650674e-05,
+      "loss": 18.4202,
+      "step": 5365
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.634046791080272e-05,
+      "loss": 16.1203,
+      "step": 5366
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.633878041088702e-05,
+      "loss": 16.4432,
+      "step": 5367
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.633709253677328e-05,
+      "loss": 18.36,
+      "step": 5368
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.633540428847517e-05,
+      "loss": 16.5339,
+      "step": 5369
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.633371566600626e-05,
+      "loss": 18.0746,
+      "step": 5370
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.633202666938023e-05,
+      "loss": 18.4865,
+      "step": 5371
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.633033729861073e-05,
+      "loss": 16.3069,
+      "step": 5372
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.632864755371137e-05,
+      "loss": 15.8611,
+      "step": 5373
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.63269574346958e-05,
+      "loss": 15.5608,
+      "step": 5374
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.632526694157769e-05,
+      "loss": 16.9785,
+      "step": 5375
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.632357607437065e-05,
+      "loss": 15.2636,
+      "step": 5376
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.632188483308837e-05,
+      "loss": 16.9455,
+      "step": 5377
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.63201932177445e-05,
+      "loss": 16.8292,
+      "step": 5378
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.631850122835268e-05,
+      "loss": 16.7681,
+      "step": 5379
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.631680886492661e-05,
+      "loss": 17.0309,
+      "step": 5380
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.631511612747994e-05,
+      "loss": 16.9053,
+      "step": 5381
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.631342301602631e-05,
+      "loss": 15.6797,
+      "step": 5382
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.631172953057944e-05,
+      "loss": 14.8583,
+      "step": 5383
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.631003567115298e-05,
+      "loss": 16.9254,
+      "step": 5384
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.63083414377606e-05,
+      "loss": 19.0823,
+      "step": 5385
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.630664683041602e-05,
+      "loss": 17.5478,
+      "step": 5386
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.630495184913286e-05,
+      "loss": 17.4864,
+      "step": 5387
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.630325649392487e-05,
+      "loss": 18.1241,
+      "step": 5388
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.63015607648057e-05,
+      "loss": 15.8607,
+      "step": 5389
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.629986466178908e-05,
+      "loss": 15.7738,
+      "step": 5390
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.629816818488866e-05,
+      "loss": 17.0697,
+      "step": 5391
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.629647133411819e-05,
+      "loss": 17.314,
+      "step": 5392
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.629477410949132e-05,
+      "loss": 18.9859,
+      "step": 5393
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.629307651102181e-05,
+      "loss": 16.8789,
+      "step": 5394
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.629137853872332e-05,
+      "loss": 15.2013,
+      "step": 5395
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.628968019260959e-05,
+      "loss": 16.4061,
+      "step": 5396
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.628798147269433e-05,
+      "loss": 15.3151,
+      "step": 5397
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.628628237899126e-05,
+      "loss": 15.4778,
+      "step": 5398
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.62845829115141e-05,
+      "loss": 16.6516,
+      "step": 5399
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.628288307027655e-05,
+      "loss": 15.8118,
+      "step": 5400
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.628118285529237e-05,
+      "loss": 19.0115,
+      "step": 5401
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.627948226657526e-05,
+      "loss": 17.382,
+      "step": 5402
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.627778130413898e-05,
+      "loss": 19.083,
+      "step": 5403
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.627607996799726e-05,
+      "loss": 16.6066,
+      "step": 5404
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.627437825816384e-05,
+      "loss": 15.9869,
+      "step": 5405
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.627267617465243e-05,
+      "loss": 16.0367,
+      "step": 5406
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.627097371747682e-05,
+      "loss": 16.769,
+      "step": 5407
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.626927088665073e-05,
+      "loss": 16.4722,
+      "step": 5408
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.626756768218791e-05,
+      "loss": 17.781,
+      "step": 5409
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.626586410410212e-05,
+      "loss": 18.048,
+      "step": 5410
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.626416015240714e-05,
+      "loss": 17.4767,
+      "step": 5411
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.626245582711669e-05,
+      "loss": 14.4172,
+      "step": 5412
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.626075112824457e-05,
+      "loss": 17.508,
+      "step": 5413
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.625904605580452e-05,
+      "loss": 17.6026,
+      "step": 5414
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.625734060981029e-05,
+      "loss": 16.9783,
+      "step": 5415
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.62556347902757e-05,
+      "loss": 16.9582,
+      "step": 5416
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.625392859721452e-05,
+      "loss": 17.6556,
+      "step": 5417
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.625222203064048e-05,
+      "loss": 16.4927,
+      "step": 5418
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.625051509056742e-05,
+      "loss": 16.756,
+      "step": 5419
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.624880777700909e-05,
+      "loss": 14.7135,
+      "step": 5420
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.624710008997926e-05,
+      "loss": 18.6539,
+      "step": 5421
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.624539202949177e-05,
+      "loss": 16.7726,
+      "step": 5422
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.624368359556037e-05,
+      "loss": 16.2563,
+      "step": 5423
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.624197478819888e-05,
+      "loss": 17.3595,
+      "step": 5424
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.624026560742108e-05,
+      "loss": 16.8117,
+      "step": 5425
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.62385560532408e-05,
+      "loss": 16.5256,
+      "step": 5426
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.623684612567184e-05,
+      "loss": 15.4597,
+      "step": 5427
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.623513582472796e-05,
+      "loss": 15.2751,
+      "step": 5428
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.623342515042303e-05,
+      "loss": 16.5878,
+      "step": 5429
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.623171410277083e-05,
+      "loss": 15.9474,
+      "step": 5430
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.623000268178522e-05,
+      "loss": 16.6678,
+      "step": 5431
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.622829088747996e-05,
+      "loss": 15.0875,
+      "step": 5432
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.622657871986891e-05,
+      "loss": 14.2974,
+      "step": 5433
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.622486617896589e-05,
+      "loss": 16.8897,
+      "step": 5434
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.622315326478472e-05,
+      "loss": 15.9572,
+      "step": 5435
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.622143997733924e-05,
+      "loss": 15.2578,
+      "step": 5436
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.62197263166433e-05,
+      "loss": 16.7446,
+      "step": 5437
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.621801228271073e-05,
+      "loss": 15.6777,
+      "step": 5438
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.621629787555533e-05,
+      "loss": 16.8873,
+      "step": 5439
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.6214583095191e-05,
+      "loss": 17.6022,
+      "step": 5440
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.621286794163158e-05,
+      "loss": 18.2273,
+      "step": 5441
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.621115241489089e-05,
+      "loss": 16.6391,
+      "step": 5442
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.62094365149828e-05,
+      "loss": 17.7344,
+      "step": 5443
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.620772024192117e-05,
+      "loss": 17.4338,
+      "step": 5444
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.620600359571987e-05,
+      "loss": 16.6578,
+      "step": 5445
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.620428657639273e-05,
+      "loss": 16.411,
+      "step": 5446
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.620256918395364e-05,
+      "loss": 18.495,
+      "step": 5447
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.620085141841646e-05,
+      "loss": 15.3195,
+      "step": 5448
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.619913327979506e-05,
+      "loss": 17.097,
+      "step": 5449
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.619741476810333e-05,
+      "loss": 15.8893,
+      "step": 5450
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.619569588335513e-05,
+      "loss": 17.0682,
+      "step": 5451
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.619397662556435e-05,
+      "loss": 15.3338,
+      "step": 5452
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.619225699474485e-05,
+      "loss": 15.2048,
+      "step": 5453
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.619053699091056e-05,
+      "loss": 17.084,
+      "step": 5454
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.618881661407533e-05,
+      "loss": 16.2449,
+      "step": 5455
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.618709586425308e-05,
+      "loss": 16.6796,
+      "step": 5456
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.618537474145767e-05,
+      "loss": 16.5597,
+      "step": 5457
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.618365324570305e-05,
+      "loss": 18.7127,
+      "step": 5458
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.618193137700309e-05,
+      "loss": 16.7265,
+      "step": 5459
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.618020913537169e-05,
+      "loss": 16.504,
+      "step": 5460
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.617848652082275e-05,
+      "loss": 15.1657,
+      "step": 5461
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.617676353337022e-05,
+      "loss": 16.8565,
+      "step": 5462
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.617504017302798e-05,
+      "loss": 15.2998,
+      "step": 5463
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.617331643980996e-05,
+      "loss": 18.089,
+      "step": 5464
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.617159233373008e-05,
+      "loss": 17.403,
+      "step": 5465
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.616986785480224e-05,
+      "loss": 16.0094,
+      "step": 5466
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.61681430030404e-05,
+      "loss": 15.8115,
+      "step": 5467
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.616641777845848e-05,
+      "loss": 15.3583,
+      "step": 5468
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.616469218107038e-05,
+      "loss": 15.4918,
+      "step": 5469
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.616296621089008e-05,
+      "loss": 15.8518,
+      "step": 5470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.616123986793148e-05,
+      "loss": 16.8515,
+      "step": 5471
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.615951315220856e-05,
+      "loss": 15.7108,
+      "step": 5472
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.615778606373523e-05,
+      "loss": 15.0067,
+      "step": 5473
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.615605860252544e-05,
+      "loss": 15.3679,
+      "step": 5474
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.615433076859315e-05,
+      "loss": 16.0537,
+      "step": 5475
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.615260256195233e-05,
+      "loss": 17.2949,
+      "step": 5476
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.61508739826169e-05,
+      "loss": 16.9238,
+      "step": 5477
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.614914503060083e-05,
+      "loss": 17.3123,
+      "step": 5478
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.614741570591809e-05,
+      "loss": 16.2074,
+      "step": 5479
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.614568600858264e-05,
+      "loss": 17.1478,
+      "step": 5480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.614395593860845e-05,
+      "loss": 17.0379,
+      "step": 5481
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.61422254960095e-05,
+      "loss": 17.3336,
+      "step": 5482
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.614049468079974e-05,
+      "loss": 15.1927,
+      "step": 5483
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.613876349299317e-05,
+      "loss": 16.356,
+      "step": 5484
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.613703193260376e-05,
+      "loss": 16.4073,
+      "step": 5485
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.613529999964547e-05,
+      "loss": 16.9577,
+      "step": 5486
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.613356769413234e-05,
+      "loss": 14.7292,
+      "step": 5487
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.613183501607831e-05,
+      "loss": 16.5073,
+      "step": 5488
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.613010196549741e-05,
+      "loss": 17.0531,
+      "step": 5489
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.612836854240358e-05,
+      "loss": 16.4902,
+      "step": 5490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.612663474681088e-05,
+      "loss": 17.247,
+      "step": 5491
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.612490057873328e-05,
+      "loss": 15.8567,
+      "step": 5492
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.612316603818479e-05,
+      "loss": 17.3632,
+      "step": 5493
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.612143112517939e-05,
+      "loss": 15.3739,
+      "step": 5494
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.611969583973114e-05,
+      "loss": 15.7259,
+      "step": 5495
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.611796018185402e-05,
+      "loss": 17.0155,
+      "step": 5496
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.611622415156204e-05,
+      "loss": 16.9418,
+      "step": 5497
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.611448774886924e-05,
+      "loss": 14.5774,
+      "step": 5498
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.611275097378963e-05,
+      "loss": 15.6845,
+      "step": 5499
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.611101382633725e-05,
+      "loss": 15.5325,
+      "step": 5500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.61092763065261e-05,
+      "loss": 18.2465,
+      "step": 5501
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.610753841437024e-05,
+      "loss": 14.5682,
+      "step": 5502
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.610580014988369e-05,
+      "loss": 16.2798,
+      "step": 5503
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.610406151308046e-05,
+      "loss": 16.7902,
+      "step": 5504
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.610232250397467e-05,
+      "loss": 15.9294,
+      "step": 5505
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.610058312258028e-05,
+      "loss": 17.7039,
+      "step": 5506
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.609884336891136e-05,
+      "loss": 15.4672,
+      "step": 5507
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.609710324298198e-05,
+      "loss": 18.2282,
+      "step": 5508
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.609536274480618e-05,
+      "loss": 17.2976,
+      "step": 5509
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.609362187439798e-05,
+      "loss": 17.1989,
+      "step": 5510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.609188063177151e-05,
+      "loss": 16.0617,
+      "step": 5511
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.609013901694078e-05,
+      "loss": 18.7813,
+      "step": 5512
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.608839702991986e-05,
+      "loss": 16.6292,
+      "step": 5513
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.608665467072283e-05,
+      "loss": 16.1779,
+      "step": 5514
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.608491193936376e-05,
+      "loss": 15.9802,
+      "step": 5515
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.60831688358567e-05,
+      "loss": 15.531,
+      "step": 5516
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.608142536021575e-05,
+      "loss": 18.7168,
+      "step": 5517
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.607968151245498e-05,
+      "loss": 16.1508,
+      "step": 5518
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.607793729258847e-05,
+      "loss": 18.0499,
+      "step": 5519
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.607619270063031e-05,
+      "loss": 13.6971,
+      "step": 5520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.607444773659459e-05,
+      "loss": 15.585,
+      "step": 5521
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.607270240049539e-05,
+      "loss": 15.9559,
+      "step": 5522
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.607095669234681e-05,
+      "loss": 16.5064,
+      "step": 5523
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.606921061216295e-05,
+      "loss": 16.345,
+      "step": 5524
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.60674641599579e-05,
+      "loss": 15.6491,
+      "step": 5525
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.60657173357458e-05,
+      "loss": 15.5191,
+      "step": 5526
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.606397013954071e-05,
+      "loss": 15.9844,
+      "step": 5527
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.606222257135675e-05,
+      "loss": 16.1229,
+      "step": 5528
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.606047463120805e-05,
+      "loss": 15.8852,
+      "step": 5529
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.60587263191087e-05,
+      "loss": 17.5455,
+      "step": 5530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.605697763507285e-05,
+      "loss": 16.3021,
+      "step": 5531
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.60552285791146e-05,
+      "loss": 16.139,
+      "step": 5532
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.605347915124807e-05,
+      "loss": 16.7253,
+      "step": 5533
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.605172935148739e-05,
+      "loss": 17.5822,
+      "step": 5534
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.604997917984669e-05,
+      "loss": 16.456,
+      "step": 5535
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.604822863634013e-05,
+      "loss": 16.3105,
+      "step": 5536
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.60464777209818e-05,
+      "loss": 16.8773,
+      "step": 5537
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.604472643378586e-05,
+      "loss": 16.0328,
+      "step": 5538
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.604297477476647e-05,
+      "loss": 15.5464,
+      "step": 5539
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.604122274393777e-05,
+      "loss": 18.6151,
+      "step": 5540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.603947034131388e-05,
+      "loss": 15.4664,
+      "step": 5541
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.603771756690897e-05,
+      "loss": 16.5297,
+      "step": 5542
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.603596442073718e-05,
+      "loss": 17.5109,
+      "step": 5543
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.60342109028127e-05,
+      "loss": 16.0709,
+      "step": 5544
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.603245701314967e-05,
+      "loss": 15.6761,
+      "step": 5545
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.603070275176225e-05,
+      "loss": 17.6334,
+      "step": 5546
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.60289481186646e-05,
+      "loss": 17.3119,
+      "step": 5547
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.602719311387092e-05,
+      "loss": 15.6359,
+      "step": 5548
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.602543773739535e-05,
+      "loss": 17.1697,
+      "step": 5549
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.602368198925207e-05,
+      "loss": 15.0955,
+      "step": 5550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.602192586945528e-05,
+      "loss": 15.967,
+      "step": 5551
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.602016937801912e-05,
+      "loss": 16.421,
+      "step": 5552
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.601841251495782e-05,
+      "loss": 15.0666,
+      "step": 5553
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.601665528028554e-05,
+      "loss": 15.9086,
+      "step": 5554
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.601489767401648e-05,
+      "loss": 15.7919,
+      "step": 5555
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.601313969616482e-05,
+      "loss": 17.7206,
+      "step": 5556
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.601138134674478e-05,
+      "loss": 15.6992,
+      "step": 5557
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.600962262577053e-05,
+      "loss": 15.3461,
+      "step": 5558
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.60078635332563e-05,
+      "loss": 17.3707,
+      "step": 5559
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.600610406921627e-05,
+      "loss": 16.6714,
+      "step": 5560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.600434423366467e-05,
+      "loss": 16.8961,
+      "step": 5561
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.60025840266157e-05,
+      "loss": 18.8765,
+      "step": 5562
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.600082344808357e-05,
+      "loss": 17.4021,
+      "step": 5563
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.59990624980825e-05,
+      "loss": 14.385,
+      "step": 5564
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.599730117662672e-05,
+      "loss": 17.8724,
+      "step": 5565
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.599553948373045e-05,
+      "loss": 15.7385,
+      "step": 5566
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.599377741940792e-05,
+      "loss": 14.4359,
+      "step": 5567
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.599201498367334e-05,
+      "loss": 17.8474,
+      "step": 5568
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.599025217654096e-05,
+      "loss": 16.3484,
+      "step": 5569
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.598848899802499e-05,
+      "loss": 15.8071,
+      "step": 5570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.59867254481397e-05,
+      "loss": 16.9404,
+      "step": 5571
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.598496152689933e-05,
+      "loss": 16.0619,
+      "step": 5572
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.59831972343181e-05,
+      "loss": 16.6918,
+      "step": 5573
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.598143257041025e-05,
+      "loss": 18.2838,
+      "step": 5574
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.597966753519007e-05,
+      "loss": 14.7351,
+      "step": 5575
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.59779021286718e-05,
+      "loss": 15.6764,
+      "step": 5576
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.597613635086969e-05,
+      "loss": 16.7157,
+      "step": 5577
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.597437020179798e-05,
+      "loss": 15.6078,
+      "step": 5578
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.597260368147096e-05,
+      "loss": 16.6257,
+      "step": 5579
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.597083678990289e-05,
+      "loss": 17.777,
+      "step": 5580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.596906952710802e-05,
+      "loss": 18.2911,
+      "step": 5581
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.596730189310063e-05,
+      "loss": 16.1475,
+      "step": 5582
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.596553388789501e-05,
+      "loss": 17.5153,
+      "step": 5583
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.596376551150544e-05,
+      "loss": 14.891,
+      "step": 5584
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.596199676394616e-05,
+      "loss": 18.1438,
+      "step": 5585
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.596022764523149e-05,
+      "loss": 15.7658,
+      "step": 5586
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.59584581553757e-05,
+      "loss": 17.3441,
+      "step": 5587
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.595668829439308e-05,
+      "loss": 15.7206,
+      "step": 5588
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.595491806229795e-05,
+      "loss": 17.1176,
+      "step": 5589
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.595314745910456e-05,
+      "loss": 16.9452,
+      "step": 5590
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.595137648482724e-05,
+      "loss": 14.8045,
+      "step": 5591
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.594960513948027e-05,
+      "loss": 15.7245,
+      "step": 5592
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.594783342307795e-05,
+      "loss": 18.0419,
+      "step": 5593
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.594606133563464e-05,
+      "loss": 17.5076,
+      "step": 5594
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.59442888771646e-05,
+      "loss": 16.2391,
+      "step": 5595
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.594251604768214e-05,
+      "loss": 17.7758,
+      "step": 5596
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.594074284720159e-05,
+      "loss": 15.2632,
+      "step": 5597
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.593896927573728e-05,
+      "loss": 15.6732,
+      "step": 5598
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.593719533330351e-05,
+      "loss": 16.9622,
+      "step": 5599
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.593542101991465e-05,
+      "loss": 16.9708,
+      "step": 5600
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.593364633558497e-05,
+      "loss": 18.1454,
+      "step": 5601
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.593187128032881e-05,
+      "loss": 18.274,
+      "step": 5602
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.593009585416054e-05,
+      "loss": 16.8753,
+      "step": 5603
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.592832005709448e-05,
+      "loss": 18.3956,
+      "step": 5604
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.592654388914495e-05,
+      "loss": 16.0022,
+      "step": 5605
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.592476735032633e-05,
+      "loss": 15.5106,
+      "step": 5606
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.592299044065294e-05,
+      "loss": 16.5648,
+      "step": 5607
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.592121316013914e-05,
+      "loss": 17.4776,
+      "step": 5608
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.591943550879926e-05,
+      "loss": 17.6182,
+      "step": 5609
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.591765748664768e-05,
+      "loss": 16.5616,
+      "step": 5610
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.591587909369876e-05,
+      "loss": 17.2572,
+      "step": 5611
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.591410032996683e-05,
+      "loss": 16.8757,
+      "step": 5612
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.59123211954663e-05,
+      "loss": 17.0858,
+      "step": 5613
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.59105416902115e-05,
+      "loss": 16.774,
+      "step": 5614
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.590876181421682e-05,
+      "loss": 17.5684,
+      "step": 5615
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.590698156749662e-05,
+      "loss": 14.0737,
+      "step": 5616
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.590520095006527e-05,
+      "loss": 15.8181,
+      "step": 5617
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.590341996193718e-05,
+      "loss": 15.793,
+      "step": 5618
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.59016386031267e-05,
+      "loss": 16.4095,
+      "step": 5619
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.589985687364823e-05,
+      "loss": 17.2979,
+      "step": 5620
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.589807477351615e-05,
+      "loss": 17.1674,
+      "step": 5621
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.589629230274487e-05,
+      "loss": 16.1009,
+      "step": 5622
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.589450946134875e-05,
+      "loss": 15.9301,
+      "step": 5623
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.589272624934222e-05,
+      "loss": 16.0427,
+      "step": 5624
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.589094266673966e-05,
+      "loss": 16.5027,
+      "step": 5625
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.588915871355548e-05,
+      "loss": 14.6392,
+      "step": 5626
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.58873743898041e-05,
+      "loss": 18.3407,
+      "step": 5627
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.58855896954999e-05,
+      "loss": 16.2306,
+      "step": 5628
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.588380463065731e-05,
+      "loss": 15.4227,
+      "step": 5629
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.588201919529074e-05,
+      "loss": 16.5563,
+      "step": 5630
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.58802333894146e-05,
+      "loss": 15.2149,
+      "step": 5631
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.587844721304334e-05,
+      "loss": 17.1721,
+      "step": 5632
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.587666066619135e-05,
+      "loss": 16.7742,
+      "step": 5633
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.587487374887307e-05,
+      "loss": 19.4408,
+      "step": 5634
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.587308646110295e-05,
+      "loss": 16.81,
+      "step": 5635
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.587129880289538e-05,
+      "loss": 15.1881,
+      "step": 5636
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.586951077426483e-05,
+      "loss": 16.5118,
+      "step": 5637
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.586772237522573e-05,
+      "loss": 17.133,
+      "step": 5638
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.58659336057925e-05,
+      "loss": 16.1478,
+      "step": 5639
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.586414446597963e-05,
+      "loss": 14.7051,
+      "step": 5640
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.586235495580151e-05,
+      "loss": 17.1036,
+      "step": 5641
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.586056507527266e-05,
+      "loss": 18.1711,
+      "step": 5642
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.585877482440748e-05,
+      "loss": 16.8511,
+      "step": 5643
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.585698420322044e-05,
+      "loss": 14.6186,
+      "step": 5644
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.5855193211726e-05,
+      "loss": 18.4609,
+      "step": 5645
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.585340184993864e-05,
+      "loss": 16.6637,
+      "step": 5646
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.585161011787281e-05,
+      "loss": 17.5899,
+      "step": 5647
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.584981801554297e-05,
+      "loss": 17.2566,
+      "step": 5648
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.58480255429636e-05,
+      "loss": 15.907,
+      "step": 5649
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.584623270014918e-05,
+      "loss": 16.2392,
+      "step": 5650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.58444394871142e-05,
+      "loss": 16.1977,
+      "step": 5651
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.58426459038731e-05,
+      "loss": 15.8055,
+      "step": 5652
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.584085195044042e-05,
+      "loss": 16.4669,
+      "step": 5653
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.583905762683058e-05,
+      "loss": 15.3765,
+      "step": 5654
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.583726293305814e-05,
+      "loss": 18.0929,
+      "step": 5655
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.583546786913754e-05,
+      "loss": 16.6469,
+      "step": 5656
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.58336724350833e-05,
+      "loss": 18.0508,
+      "step": 5657
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.58318766309099e-05,
+      "loss": 17.5725,
+      "step": 5658
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.583008045663187e-05,
+      "loss": 15.7139,
+      "step": 5659
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.582828391226369e-05,
+      "loss": 15.6706,
+      "step": 5660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.582648699781987e-05,
+      "loss": 17.1781,
+      "step": 5661
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.582468971331491e-05,
+      "loss": 15.1249,
+      "step": 5662
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.582289205876337e-05,
+      "loss": 16.7368,
+      "step": 5663
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.582109403417972e-05,
+      "loss": 16.0502,
+      "step": 5664
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.58192956395785e-05,
+      "loss": 16.3616,
+      "step": 5665
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.581749687497424e-05,
+      "loss": 14.5704,
+      "step": 5666
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.581569774038144e-05,
+      "loss": 16.5832,
+      "step": 5667
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.581389823581465e-05,
+      "loss": 17.3185,
+      "step": 5668
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.581209836128839e-05,
+      "loss": 16.8341,
+      "step": 5669
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.581029811681721e-05,
+      "loss": 17.4521,
+      "step": 5670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.580849750241562e-05,
+      "loss": 15.7723,
+      "step": 5671
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.580669651809817e-05,
+      "loss": 18.0151,
+      "step": 5672
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.580489516387943e-05,
+      "loss": 16.2284,
+      "step": 5673
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.580309343977392e-05,
+      "loss": 17.7795,
+      "step": 5674
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.580129134579619e-05,
+      "loss": 15.4943,
+      "step": 5675
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.57994888819608e-05,
+      "loss": 17.3132,
+      "step": 5676
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.579768604828232e-05,
+      "loss": 19.4052,
+      "step": 5677
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.579588284477526e-05,
+      "loss": 15.1527,
+      "step": 5678
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.579407927145423e-05,
+      "loss": 16.1703,
+      "step": 5679
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.579227532833377e-05,
+      "loss": 14.9781,
+      "step": 5680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.579047101542847e-05,
+      "loss": 17.7104,
+      "step": 5681
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.578866633275288e-05,
+      "loss": 17.619,
+      "step": 5682
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.578686128032157e-05,
+      "loss": 16.5793,
+      "step": 5683
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.578505585814913e-05,
+      "loss": 16.4968,
+      "step": 5684
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.578325006625012e-05,
+      "loss": 17.1128,
+      "step": 5685
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.578144390463917e-05,
+      "loss": 18.257,
+      "step": 5686
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.57796373733308e-05,
+      "loss": 16.7381,
+      "step": 5687
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.577783047233963e-05,
+      "loss": 17.1365,
+      "step": 5688
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.577602320168028e-05,
+      "loss": 18.0327,
+      "step": 5689
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.57742155613673e-05,
+      "loss": 15.4846,
+      "step": 5690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.57724075514153e-05,
+      "loss": 15.6443,
+      "step": 5691
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.577059917183885e-05,
+      "loss": 16.1008,
+      "step": 5692
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.576879042265263e-05,
+      "loss": 15.5796,
+      "step": 5693
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.576698130387119e-05,
+      "loss": 17.1051,
+      "step": 5694
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.576517181550914e-05,
+      "loss": 15.6703,
+      "step": 5695
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.57633619575811e-05,
+      "loss": 16.1438,
+      "step": 5696
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.576155173010172e-05,
+      "loss": 16.56,
+      "step": 5697
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.575974113308557e-05,
+      "loss": 17.0058,
+      "step": 5698
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.575793016654727e-05,
+      "loss": 16.9333,
+      "step": 5699
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.575611883050148e-05,
+      "loss": 18.4461,
+      "step": 5700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.575430712496281e-05,
+      "loss": 18.8484,
+      "step": 5701
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.575249504994586e-05,
+      "loss": 15.5326,
+      "step": 5702
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.575068260546531e-05,
+      "loss": 15.8163,
+      "step": 5703
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.574886979153578e-05,
+      "loss": 16.4325,
+      "step": 5704
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.57470566081719e-05,
+      "loss": 17.1178,
+      "step": 5705
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.574524305538833e-05,
+      "loss": 16.1212,
+      "step": 5706
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.574342913319969e-05,
+      "loss": 17.28,
+      "step": 5707
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.574161484162064e-05,
+      "loss": 15.1332,
+      "step": 5708
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.573980018066584e-05,
+      "loss": 15.6723,
+      "step": 5709
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.573798515034992e-05,
+      "loss": 14.4106,
+      "step": 5710
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.573616975068758e-05,
+      "loss": 16.2168,
+      "step": 5711
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.573435398169343e-05,
+      "loss": 16.7962,
+      "step": 5712
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.573253784338216e-05,
+      "loss": 14.4034,
+      "step": 5713
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.573072133576844e-05,
+      "loss": 18.1536,
+      "step": 5714
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.572890445886691e-05,
+      "loss": 16.7966,
+      "step": 5715
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.572708721269228e-05,
+      "loss": 17.9185,
+      "step": 5716
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.57252695972592e-05,
+      "loss": 18.6677,
+      "step": 5717
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.572345161258235e-05,
+      "loss": 16.0234,
+      "step": 5718
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.572163325867643e-05,
+      "loss": 17.1275,
+      "step": 5719
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.571981453555608e-05,
+      "loss": 16.8561,
+      "step": 5720
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.571799544323602e-05,
+      "loss": 16.2973,
+      "step": 5721
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.571617598173095e-05,
+      "loss": 17.0625,
+      "step": 5722
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.571435615105555e-05,
+      "loss": 16.8299,
+      "step": 5723
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.571253595122452e-05,
+      "loss": 19.3813,
+      "step": 5724
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.571071538225253e-05,
+      "loss": 18.1197,
+      "step": 5725
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.570889444415431e-05,
+      "loss": 16.5072,
+      "step": 5726
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.570707313694456e-05,
+      "loss": 15.9835,
+      "step": 5727
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.570525146063798e-05,
+      "loss": 15.635,
+      "step": 5728
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.570342941524928e-05,
+      "loss": 17.2323,
+      "step": 5729
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.570160700079319e-05,
+      "loss": 16.2489,
+      "step": 5730
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.569978421728442e-05,
+      "loss": 16.9828,
+      "step": 5731
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.569796106473767e-05,
+      "loss": 18.9131,
+      "step": 5732
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.569613754316769e-05,
+      "loss": 16.7216,
+      "step": 5733
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.569431365258918e-05,
+      "loss": 17.1829,
+      "step": 5734
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.569248939301689e-05,
+      "loss": 18.752,
+      "step": 5735
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.569066476446555e-05,
+      "loss": 16.3422,
+      "step": 5736
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.568883976694989e-05,
+      "loss": 16.3107,
+      "step": 5737
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.568701440048462e-05,
+      "loss": 18.1053,
+      "step": 5738
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.568518866508452e-05,
+      "loss": 16.0469,
+      "step": 5739
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.568336256076433e-05,
+      "loss": 16.4656,
+      "step": 5740
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.568153608753876e-05,
+      "loss": 17.378,
+      "step": 5741
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.56797092454226e-05,
+      "loss": 14.5834,
+      "step": 5742
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.567788203443057e-05,
+      "loss": 18.2215,
+      "step": 5743
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.567605445457747e-05,
+      "loss": 17.2549,
+      "step": 5744
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.567422650587801e-05,
+      "loss": 15.2247,
+      "step": 5745
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.567239818834696e-05,
+      "loss": 17.2839,
+      "step": 5746
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.567056950199912e-05,
+      "loss": 15.8446,
+      "step": 5747
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.566874044684921e-05,
+      "loss": 15.8676,
+      "step": 5748
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.566691102291203e-05,
+      "loss": 16.3216,
+      "step": 5749
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.566508123020235e-05,
+      "loss": 16.7006,
+      "step": 5750
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.566325106873494e-05,
+      "loss": 15.8988,
+      "step": 5751
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.566142053852458e-05,
+      "loss": 17.0842,
+      "step": 5752
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.565958963958605e-05,
+      "loss": 17.7166,
+      "step": 5753
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.565775837193413e-05,
+      "loss": 16.256,
+      "step": 5754
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.565592673558362e-05,
+      "loss": 15.4604,
+      "step": 5755
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.565409473054932e-05,
+      "loss": 16.7319,
+      "step": 5756
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.565226235684599e-05,
+      "loss": 15.7891,
+      "step": 5757
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.565042961448844e-05,
+      "loss": 16.4941,
+      "step": 5758
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.564859650349149e-05,
+      "loss": 16.8651,
+      "step": 5759
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.564676302386994e-05,
+      "loss": 16.7461,
+      "step": 5760
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.564492917563858e-05,
+      "loss": 13.5616,
+      "step": 5761
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.564309495881221e-05,
+      "loss": 14.9247,
+      "step": 5762
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.564126037340567e-05,
+      "loss": 15.2606,
+      "step": 5763
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.563942541943375e-05,
+      "loss": 16.6577,
+      "step": 5764
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.56375900969113e-05,
+      "loss": 13.0682,
+      "step": 5765
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.56357544058531e-05,
+      "loss": 17.046,
+      "step": 5766
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.5633918346274e-05,
+      "loss": 15.7223,
+      "step": 5767
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.563208191818881e-05,
+      "loss": 17.8095,
+      "step": 5768
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.563024512161238e-05,
+      "loss": 15.9073,
+      "step": 5769
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.562840795655951e-05,
+      "loss": 17.2289,
+      "step": 5770
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.56265704230451e-05,
+      "loss": 15.6762,
+      "step": 5771
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.56247325210839e-05,
+      "loss": 16.9419,
+      "step": 5772
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.562289425069083e-05,
+      "loss": 17.8811,
+      "step": 5773
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.562105561188069e-05,
+      "loss": 16.9909,
+      "step": 5774
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.561921660466835e-05,
+      "loss": 16.6802,
+      "step": 5775
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.561737722906862e-05,
+      "loss": 17.6176,
+      "step": 5776
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.561553748509641e-05,
+      "loss": 16.8329,
+      "step": 5777
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.561369737276656e-05,
+      "loss": 15.2225,
+      "step": 5778
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.561185689209389e-05,
+      "loss": 17.5244,
+      "step": 5779
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.561001604309331e-05,
+      "loss": 15.0868,
+      "step": 5780
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.560817482577966e-05,
+      "loss": 16.4401,
+      "step": 5781
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.560633324016782e-05,
+      "loss": 17.5649,
+      "step": 5782
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.560449128627265e-05,
+      "loss": 17.2878,
+      "step": 5783
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.560264896410904e-05,
+      "loss": 15.7453,
+      "step": 5784
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.560080627369186e-05,
+      "loss": 16.1921,
+      "step": 5785
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.559896321503599e-05,
+      "loss": 16.6708,
+      "step": 5786
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.55971197881563e-05,
+      "loss": 15.8001,
+      "step": 5787
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.55952759930677e-05,
+      "loss": 15.9287,
+      "step": 5788
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.559343182978504e-05,
+      "loss": 16.4954,
+      "step": 5789
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.559158729832327e-05,
+      "loss": 17.2574,
+      "step": 5790
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.558974239869724e-05,
+      "loss": 16.0001,
+      "step": 5791
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.558789713092187e-05,
+      "loss": 16.5735,
+      "step": 5792
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.558605149501205e-05,
+      "loss": 15.7325,
+      "step": 5793
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.558420549098268e-05,
+      "loss": 16.6778,
+      "step": 5794
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.558235911884868e-05,
+      "loss": 16.5266,
+      "step": 5795
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.558051237862495e-05,
+      "loss": 17.8094,
+      "step": 5796
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.55786652703264e-05,
+      "loss": 16.744,
+      "step": 5797
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.557681779396797e-05,
+      "loss": 16.506,
+      "step": 5798
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.557496994956456e-05,
+      "loss": 17.4896,
+      "step": 5799
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.557312173713109e-05,
+      "loss": 15.4937,
+      "step": 5800
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.557127315668248e-05,
+      "loss": 15.0682,
+      "step": 5801
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.556942420823368e-05,
+      "loss": 17.0604,
+      "step": 5802
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.55675748917996e-05,
+      "loss": 16.3869,
+      "step": 5803
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.556572520739518e-05,
+      "loss": 15.2857,
+      "step": 5804
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.556387515503535e-05,
+      "loss": 17.2385,
+      "step": 5805
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.556202473473508e-05,
+      "loss": 15.0139,
+      "step": 5806
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.556017394650927e-05,
+      "loss": 18.5064,
+      "step": 5807
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.555832279037289e-05,
+      "loss": 18.1776,
+      "step": 5808
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.555647126634088e-05,
+      "loss": 18.0807,
+      "step": 5809
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.55546193744282e-05,
+      "loss": 15.9956,
+      "step": 5810
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.555276711464979e-05,
+      "loss": 15.9483,
+      "step": 5811
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.555091448702065e-05,
+      "loss": 16.6575,
+      "step": 5812
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.554906149155568e-05,
+      "loss": 18.3671,
+      "step": 5813
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.554720812826987e-05,
+      "loss": 15.9602,
+      "step": 5814
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.55453543971782e-05,
+      "loss": 16.2569,
+      "step": 5815
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.554350029829562e-05,
+      "loss": 16.0801,
+      "step": 5816
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.554164583163711e-05,
+      "loss": 15.9006,
+      "step": 5817
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.553979099721766e-05,
+      "loss": 19.2141,
+      "step": 5818
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.553793579505221e-05,
+      "loss": 15.4901,
+      "step": 5819
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.553608022515577e-05,
+      "loss": 16.4606,
+      "step": 5820
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.553422428754333e-05,
+      "loss": 16.2528,
+      "step": 5821
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.553236798222986e-05,
+      "loss": 16.5541,
+      "step": 5822
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.553051130923034e-05,
+      "loss": 15.5261,
+      "step": 5823
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.55286542685598e-05,
+      "loss": 16.9041,
+      "step": 5824
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.552679686023321e-05,
+      "loss": 15.7524,
+      "step": 5825
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.552493908426558e-05,
+      "loss": 15.6219,
+      "step": 5826
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.552308094067189e-05,
+      "loss": 15.8995,
+      "step": 5827
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.552122242946717e-05,
+      "loss": 15.7764,
+      "step": 5828
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.551936355066642e-05,
+      "loss": 16.4163,
+      "step": 5829
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.551750430428465e-05,
+      "loss": 16.9469,
+      "step": 5830
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.551564469033688e-05,
+      "loss": 16.6383,
+      "step": 5831
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.551378470883812e-05,
+      "loss": 18.0189,
+      "step": 5832
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.551192435980338e-05,
+      "loss": 16.6458,
+      "step": 5833
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.551006364324771e-05,
+      "loss": 16.4467,
+      "step": 5834
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.55082025591861e-05,
+      "loss": 15.0178,
+      "step": 5835
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.550634110763361e-05,
+      "loss": 14.8977,
+      "step": 5836
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.550447928860525e-05,
+      "loss": 17.6142,
+      "step": 5837
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.550261710211608e-05,
+      "loss": 16.9274,
+      "step": 5838
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.55007545481811e-05,
+      "loss": 15.8706,
+      "step": 5839
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.54988916268154e-05,
+      "loss": 16.4936,
+      "step": 5840
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.549702833803398e-05,
+      "loss": 14.6726,
+      "step": 5841
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.54951646818519e-05,
+      "loss": 17.7279,
+      "step": 5842
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.549330065828421e-05,
+      "loss": 16.8778,
+      "step": 5843
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.549143626734598e-05,
+      "loss": 15.0654,
+      "step": 5844
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.548957150905225e-05,
+      "loss": 16.5965,
+      "step": 5845
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.548770638341806e-05,
+      "loss": 18.6215,
+      "step": 5846
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.548584089045851e-05,
+      "loss": 17.1391,
+      "step": 5847
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.548397503018865e-05,
+      "loss": 15.7058,
+      "step": 5848
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.548210880262352e-05,
+      "loss": 16.6424,
+      "step": 5849
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.548024220777822e-05,
+      "loss": 16.8508,
+      "step": 5850
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.547837524566783e-05,
+      "loss": 16.6265,
+      "step": 5851
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.54765079163074e-05,
+      "loss": 15.957,
+      "step": 5852
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.547464021971202e-05,
+      "loss": 16.3667,
+      "step": 5853
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.547277215589675e-05,
+      "loss": 16.942,
+      "step": 5854
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.547090372487673e-05,
+      "loss": 17.2069,
+      "step": 5855
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.546903492666702e-05,
+      "loss": 15.9137,
+      "step": 5856
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.54671657612827e-05,
+      "loss": 18.145,
+      "step": 5857
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.546529622873885e-05,
+      "loss": 16.9289,
+      "step": 5858
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.546342632905061e-05,
+      "loss": 16.0791,
+      "step": 5859
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.546155606223306e-05,
+      "loss": 16.0769,
+      "step": 5860
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.545968542830128e-05,
+      "loss": 17.796,
+      "step": 5861
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.545781442727042e-05,
+      "loss": 18.8013,
+      "step": 5862
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.545594305915555e-05,
+      "loss": 18.3564,
+      "step": 5863
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.54540713239718e-05,
+      "loss": 18.1702,
+      "step": 5864
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.545219922173429e-05,
+      "loss": 16.7227,
+      "step": 5865
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.545032675245813e-05,
+      "loss": 16.3989,
+      "step": 5866
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.544845391615844e-05,
+      "loss": 18.201,
+      "step": 5867
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.544658071285033e-05,
+      "loss": 16.581,
+      "step": 5868
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.544470714254896e-05,
+      "loss": 16.5303,
+      "step": 5869
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.544283320526943e-05,
+      "loss": 16.7049,
+      "step": 5870
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.54409589010269e-05,
+      "loss": 19.8991,
+      "step": 5871
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.543908422983648e-05,
+      "loss": 16.4882,
+      "step": 5872
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.543720919171331e-05,
+      "loss": 16.4063,
+      "step": 5873
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.543533378667253e-05,
+      "loss": 16.0966,
+      "step": 5874
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.54334580147293e-05,
+      "loss": 16.2001,
+      "step": 5875
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.543158187589877e-05,
+      "loss": 16.6928,
+      "step": 5876
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.542970537019607e-05,
+      "loss": 16.4983,
+      "step": 5877
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.542782849763637e-05,
+      "loss": 15.5646,
+      "step": 5878
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.542595125823482e-05,
+      "loss": 15.253,
+      "step": 5879
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.542407365200659e-05,
+      "loss": 16.4463,
+      "step": 5880
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.542219567896681e-05,
+      "loss": 16.6396,
+      "step": 5881
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.542031733913069e-05,
+      "loss": 14.0302,
+      "step": 5882
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.541843863251335e-05,
+      "loss": 17.9497,
+      "step": 5883
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.541655955913e-05,
+      "loss": 16.4875,
+      "step": 5884
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.54146801189958e-05,
+      "loss": 16.5007,
+      "step": 5885
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.541280031212591e-05,
+      "loss": 16.2168,
+      "step": 5886
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.541092013853555e-05,
+      "loss": 16.809,
+      "step": 5887
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.540903959823988e-05,
+      "loss": 15.7694,
+      "step": 5888
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.540715869125407e-05,
+      "loss": 15.7329,
+      "step": 5889
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.540527741759333e-05,
+      "loss": 17.736,
+      "step": 5890
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.540339577727285e-05,
+      "loss": 17.0393,
+      "step": 5891
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.540151377030782e-05,
+      "loss": 14.9793,
+      "step": 5892
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.539963139671344e-05,
+      "loss": 17.329,
+      "step": 5893
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.539774865650491e-05,
+      "loss": 16.8127,
+      "step": 5894
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.539586554969742e-05,
+      "loss": 14.9112,
+      "step": 5895
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.53939820763062e-05,
+      "loss": 17.7548,
+      "step": 5896
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.539209823634645e-05,
+      "loss": 17.0338,
+      "step": 5897
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.539021402983337e-05,
+      "loss": 16.2857,
+      "step": 5898
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.53883294567822e-05,
+      "loss": 17.4869,
+      "step": 5899
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.538644451720814e-05,
+      "loss": 16.7382,
+      "step": 5900
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.538455921112643e-05,
+      "loss": 17.5208,
+      "step": 5901
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.538267353855226e-05,
+      "loss": 15.4063,
+      "step": 5902
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.538078749950088e-05,
+      "loss": 15.8003,
+      "step": 5903
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.537890109398753e-05,
+      "loss": 16.2445,
+      "step": 5904
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.537701432202744e-05,
+      "loss": 17.699,
+      "step": 5905
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.537512718363583e-05,
+      "loss": 16.1802,
+      "step": 5906
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.537323967882794e-05,
+      "loss": 16.7316,
+      "step": 5907
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.537135180761903e-05,
+      "loss": 18.1495,
+      "step": 5908
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.536946357002431e-05,
+      "loss": 17.869,
+      "step": 5909
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.536757496605908e-05,
+      "loss": 16.0644,
+      "step": 5910
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.536568599573856e-05,
+      "loss": 15.2487,
+      "step": 5911
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.5363796659078e-05,
+      "loss": 16.8936,
+      "step": 5912
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.536190695609268e-05,
+      "loss": 15.9853,
+      "step": 5913
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.536001688679782e-05,
+      "loss": 17.1956,
+      "step": 5914
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.535812645120872e-05,
+      "loss": 16.8317,
+      "step": 5915
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.535623564934063e-05,
+      "loss": 17.9026,
+      "step": 5916
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.535434448120883e-05,
+      "loss": 16.7634,
+      "step": 5917
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.535245294682857e-05,
+      "loss": 15.9679,
+      "step": 5918
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.535056104621514e-05,
+      "loss": 17.8305,
+      "step": 5919
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.534866877938382e-05,
+      "loss": 16.3747,
+      "step": 5920
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.534677614634989e-05,
+      "loss": 15.9044,
+      "step": 5921
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.534488314712863e-05,
+      "loss": 14.5324,
+      "step": 5922
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.534298978173531e-05,
+      "loss": 17.2584,
+      "step": 5923
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.534109605018526e-05,
+      "loss": 15.4673,
+      "step": 5924
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.533920195249373e-05,
+      "loss": 19.7823,
+      "step": 5925
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.533730748867604e-05,
+      "loss": 16.0912,
+      "step": 5926
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.533541265874749e-05,
+      "loss": 16.1038,
+      "step": 5927
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.533351746272335e-05,
+      "loss": 15.9714,
+      "step": 5928
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.533162190061895e-05,
+      "loss": 15.8614,
+      "step": 5929
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.532972597244962e-05,
+      "loss": 17.196,
+      "step": 5930
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.532782967823062e-05,
+      "loss": 16.982,
+      "step": 5931
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.53259330179773e-05,
+      "loss": 17.259,
+      "step": 5932
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.532403599170495e-05,
+      "loss": 18.2379,
+      "step": 5933
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.532213859942891e-05,
+      "loss": 16.9208,
+      "step": 5934
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.53202408411645e-05,
+      "loss": 18.3385,
+      "step": 5935
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.531834271692703e-05,
+      "loss": 16.8686,
+      "step": 5936
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.531644422673185e-05,
+      "loss": 15.6003,
+      "step": 5937
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.531454537059427e-05,
+      "loss": 16.2145,
+      "step": 5938
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.531264614852963e-05,
+      "loss": 17.1762,
+      "step": 5939
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.531074656055327e-05,
+      "loss": 16.7223,
+      "step": 5940
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.530884660668053e-05,
+      "loss": 15.6493,
+      "step": 5941
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.530694628692673e-05,
+      "loss": 14.2541,
+      "step": 5942
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.530504560130726e-05,
+      "loss": 16.0582,
+      "step": 5943
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.530314454983744e-05,
+      "loss": 14.9227,
+      "step": 5944
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.530124313253264e-05,
+      "loss": 16.4863,
+      "step": 5945
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.52993413494082e-05,
+      "loss": 16.7386,
+      "step": 5946
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.529743920047944e-05,
+      "loss": 16.618,
+      "step": 5947
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.529553668576179e-05,
+      "loss": 17.9317,
+      "step": 5948
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.529363380527058e-05,
+      "loss": 15.96,
+      "step": 5949
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.529173055902116e-05,
+      "loss": 16.7196,
+      "step": 5950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.528982694702895e-05,
+      "loss": 16.457,
+      "step": 5951
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.528792296930927e-05,
+      "loss": 16.6269,
+      "step": 5952
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.528601862587751e-05,
+      "loss": 17.9161,
+      "step": 5953
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.528411391674906e-05,
+      "loss": 15.5545,
+      "step": 5954
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.52822088419393e-05,
+      "loss": 16.2678,
+      "step": 5955
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.528030340146359e-05,
+      "loss": 16.0397,
+      "step": 5956
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.527839759533734e-05,
+      "loss": 15.9073,
+      "step": 5957
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.527649142357596e-05,
+      "loss": 15.1622,
+      "step": 5958
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.527458488619478e-05,
+      "loss": 16.1536,
+      "step": 5959
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.527267798320925e-05,
+      "loss": 16.1479,
+      "step": 5960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.527077071463475e-05,
+      "loss": 17.1545,
+      "step": 5961
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.526886308048669e-05,
+      "loss": 16.6084,
+      "step": 5962
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.526695508078047e-05,
+      "loss": 16.2139,
+      "step": 5963
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.52650467155315e-05,
+      "loss": 16.0165,
+      "step": 5964
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.526313798475517e-05,
+      "loss": 16.7123,
+      "step": 5965
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.526122888846692e-05,
+      "loss": 16.6779,
+      "step": 5966
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.525931942668214e-05,
+      "loss": 16.7672,
+      "step": 5967
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.525740959941629e-05,
+      "loss": 15.7808,
+      "step": 5968
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.525549940668474e-05,
+      "loss": 17.8165,
+      "step": 5969
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.525358884850298e-05,
+      "loss": 15.8038,
+      "step": 5970
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.525167792488638e-05,
+      "loss": 18.0252,
+      "step": 5971
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.524976663585039e-05,
+      "loss": 16.8705,
+      "step": 5972
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.524785498141046e-05,
+      "loss": 16.8786,
+      "step": 5973
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.5245942961582e-05,
+      "loss": 16.9894,
+      "step": 5974
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.524403057638048e-05,
+      "loss": 15.9735,
+      "step": 5975
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.524211782582132e-05,
+      "loss": 18.2167,
+      "step": 5976
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.524020470991998e-05,
+      "loss": 16.4116,
+      "step": 5977
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.52382912286919e-05,
+      "loss": 16.0987,
+      "step": 5978
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.523637738215251e-05,
+      "loss": 17.227,
+      "step": 5979
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.523446317031733e-05,
+      "loss": 15.5692,
+      "step": 5980
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.523254859320175e-05,
+      "loss": 16.9043,
+      "step": 5981
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.523063365082126e-05,
+      "loss": 16.4872,
+      "step": 5982
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.522871834319133e-05,
+      "loss": 17.3909,
+      "step": 5983
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.522680267032742e-05,
+      "loss": 15.5318,
+      "step": 5984
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.522488663224498e-05,
+      "loss": 17.4524,
+      "step": 5985
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.522297022895952e-05,
+      "loss": 17.0975,
+      "step": 5986
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.522105346048649e-05,
+      "loss": 15.1177,
+      "step": 5987
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.521913632684136e-05,
+      "loss": 15.1339,
+      "step": 5988
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.521721882803964e-05,
+      "loss": 15.9284,
+      "step": 5989
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.521530096409679e-05,
+      "loss": 15.6705,
+      "step": 5990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.52133827350283e-05,
+      "loss": 17.2155,
+      "step": 5991
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.521146414084969e-05,
+      "loss": 15.9192,
+      "step": 5992
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.520954518157643e-05,
+      "loss": 16.172,
+      "step": 5993
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.520762585722401e-05,
+      "loss": 15.7157,
+      "step": 5994
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.520570616780791e-05,
+      "loss": 17.9646,
+      "step": 5995
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.520378611334368e-05,
+      "loss": 15.6837,
+      "step": 5996
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.52018656938468e-05,
+      "loss": 15.5055,
+      "step": 5997
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.519994490933279e-05,
+      "loss": 16.4498,
+      "step": 5998
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.519802375981714e-05,
+      "loss": 16.4977,
+      "step": 5999
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.519610224531538e-05,
+      "loss": 18.1628,
+      "step": 6000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.519418036584301e-05,
+      "loss": 18.335,
+      "step": 6001
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.519225812141556e-05,
+      "loss": 16.4105,
+      "step": 6002
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.519033551204857e-05,
+      "loss": 16.6421,
+      "step": 6003
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.518841253775755e-05,
+      "loss": 15.2424,
+      "step": 6004
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.518648919855801e-05,
+      "loss": 15.009,
+      "step": 6005
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.518456549446553e-05,
+      "loss": 18.4016,
+      "step": 6006
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.518264142549557e-05,
+      "loss": 15.2931,
+      "step": 6007
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.518071699166376e-05,
+      "loss": 15.7093,
+      "step": 6008
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.517879219298556e-05,
+      "loss": 17.6415,
+      "step": 6009
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.517686702947655e-05,
+      "loss": 16.4948,
+      "step": 6010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.517494150115228e-05,
+      "loss": 16.9427,
+      "step": 6011
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.517301560802828e-05,
+      "loss": 16.6796,
+      "step": 6012
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.517108935012012e-05,
+      "loss": 16.1977,
+      "step": 6013
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.516916272744333e-05,
+      "loss": 14.0611,
+      "step": 6014
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.51672357400135e-05,
+      "loss": 17.0108,
+      "step": 6015
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.516530838784617e-05,
+      "loss": 17.0529,
+      "step": 6016
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.516338067095691e-05,
+      "loss": 15.7255,
+      "step": 6017
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.51614525893613e-05,
+      "loss": 14.6963,
+      "step": 6018
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.515952414307489e-05,
+      "loss": 18.0499,
+      "step": 6019
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.515759533211326e-05,
+      "loss": 13.7549,
+      "step": 6020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.515566615649198e-05,
+      "loss": 16.6454,
+      "step": 6021
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.515373661622664e-05,
+      "loss": 16.8708,
+      "step": 6022
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.515180671133281e-05,
+      "loss": 16.336,
+      "step": 6023
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.514987644182609e-05,
+      "loss": 17.4496,
+      "step": 6024
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.514794580772203e-05,
+      "loss": 15.8379,
+      "step": 6025
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.51460148090363e-05,
+      "loss": 18.3696,
+      "step": 6026
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.514408344578439e-05,
+      "loss": 15.0669,
+      "step": 6027
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.514215171798197e-05,
+      "loss": 19.6574,
+      "step": 6028
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.514021962564462e-05,
+      "loss": 16.5415,
+      "step": 6029
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.513828716878792e-05,
+      "loss": 17.7048,
+      "step": 6030
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.51363543474275e-05,
+      "loss": 16.7791,
+      "step": 6031
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.513442116157897e-05,
+      "loss": 17.5791,
+      "step": 6032
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.513248761125793e-05,
+      "loss": 18.9578,
+      "step": 6033
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.513055369648e-05,
+      "loss": 13.9957,
+      "step": 6034
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.512861941726078e-05,
+      "loss": 15.0515,
+      "step": 6035
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.512668477361593e-05,
+      "loss": 15.7202,
+      "step": 6036
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.512474976556103e-05,
+      "loss": 17.5336,
+      "step": 6037
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.51228143931117e-05,
+      "loss": 14.8323,
+      "step": 6038
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.512087865628363e-05,
+      "loss": 16.5053,
+      "step": 6039
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.51189425550924e-05,
+      "loss": 16.1464,
+      "step": 6040
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.511700608955364e-05,
+      "loss": 16.0532,
+      "step": 6041
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.5115069259683e-05,
+      "loss": 15.7969,
+      "step": 6042
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.511313206549615e-05,
+      "loss": 16.0115,
+      "step": 6043
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.511119450700869e-05,
+      "loss": 14.6728,
+      "step": 6044
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.510925658423631e-05,
+      "loss": 16.5239,
+      "step": 6045
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.510731829719461e-05,
+      "loss": 14.7533,
+      "step": 6046
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.510537964589928e-05,
+      "loss": 16.0768,
+      "step": 6047
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.510344063036594e-05,
+      "loss": 14.7092,
+      "step": 6048
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.510150125061028e-05,
+      "loss": 16.1841,
+      "step": 6049
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.509956150664796e-05,
+      "loss": 16.079,
+      "step": 6050
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.509762139849462e-05,
+      "loss": 16.9618,
+      "step": 6051
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.509568092616594e-05,
+      "loss": 15.1375,
+      "step": 6052
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.509374008967759e-05,
+      "loss": 16.2047,
+      "step": 6053
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.509179888904526e-05,
+      "loss": 15.5824,
+      "step": 6054
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.508985732428459e-05,
+      "loss": 16.6504,
+      "step": 6055
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.508791539541129e-05,
+      "loss": 15.9224,
+      "step": 6056
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.508597310244102e-05,
+      "loss": 17.4529,
+      "step": 6057
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.508403044538948e-05,
+      "loss": 16.5891,
+      "step": 6058
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.508208742427237e-05,
+      "loss": 16.678,
+      "step": 6059
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.508014403910533e-05,
+      "loss": 17.2427,
+      "step": 6060
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.507820028990409e-05,
+      "loss": 14.4299,
+      "step": 6061
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.507625617668435e-05,
+      "loss": 19.0142,
+      "step": 6062
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.50743116994618e-05,
+      "loss": 17.7837,
+      "step": 6063
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.507236685825214e-05,
+      "loss": 17.8762,
+      "step": 6064
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.507042165307108e-05,
+      "loss": 15.5874,
+      "step": 6065
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.506847608393432e-05,
+      "loss": 17.7825,
+      "step": 6066
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.50665301508576e-05,
+      "loss": 18.4464,
+      "step": 6067
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.506458385385659e-05,
+      "loss": 17.0419,
+      "step": 6068
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.506263719294703e-05,
+      "loss": 16.3529,
+      "step": 6069
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.506069016814463e-05,
+      "loss": 16.7774,
+      "step": 6070
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.505874277946514e-05,
+      "loss": 16.2353,
+      "step": 6071
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.505679502692425e-05,
+      "loss": 17.2325,
+      "step": 6072
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.505484691053771e-05,
+      "loss": 15.137,
+      "step": 6073
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.505289843032124e-05,
+      "loss": 16.8955,
+      "step": 6074
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.505094958629059e-05,
+      "loss": 15.3058,
+      "step": 6075
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.504900037846148e-05,
+      "loss": 17.1881,
+      "step": 6076
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.504705080684967e-05,
+      "loss": 16.5778,
+      "step": 6077
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.504510087147088e-05,
+      "loss": 15.5334,
+      "step": 6078
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.504315057234088e-05,
+      "loss": 15.7947,
+      "step": 6079
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.50411999094754e-05,
+      "loss": 16.6895,
+      "step": 6080
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.503924888289019e-05,
+      "loss": 15.0043,
+      "step": 6081
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.503729749260102e-05,
+      "loss": 19.1661,
+      "step": 6082
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.503534573862364e-05,
+      "loss": 16.844,
+      "step": 6083
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.503339362097384e-05,
+      "loss": 17.1609,
+      "step": 6084
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.503144113966734e-05,
+      "loss": 17.2736,
+      "step": 6085
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.502948829471991e-05,
+      "loss": 15.9259,
+      "step": 6086
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.502753508614734e-05,
+      "loss": 17.5937,
+      "step": 6087
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.502558151396542e-05,
+      "loss": 17.0539,
+      "step": 6088
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.502362757818987e-05,
+      "loss": 15.6691,
+      "step": 6089
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.502167327883652e-05,
+      "loss": 16.2769,
+      "step": 6090
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.501971861592112e-05,
+      "loss": 16.3186,
+      "step": 6091
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.501776358945946e-05,
+      "loss": 14.9158,
+      "step": 6092
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.501580819946736e-05,
+      "loss": 16.9302,
+      "step": 6093
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.501385244596056e-05,
+      "loss": 17.5242,
+      "step": 6094
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.501189632895488e-05,
+      "loss": 17.0282,
+      "step": 6095
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.500993984846614e-05,
+      "loss": 15.4503,
+      "step": 6096
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.500798300451009e-05,
+      "loss": 16.4168,
+      "step": 6097
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.500602579710256e-05,
+      "loss": 15.616,
+      "step": 6098
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.500406822625933e-05,
+      "loss": 17.0344,
+      "step": 6099
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.500211029199626e-05,
+      "loss": 18.2779,
+      "step": 6100
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.500015199432912e-05,
+      "loss": 14.1762,
+      "step": 6101
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.499819333327371e-05,
+      "loss": 16.9975,
+      "step": 6102
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.499623430884588e-05,
+      "loss": 15.4109,
+      "step": 6103
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.499427492106145e-05,
+      "loss": 16.0809,
+      "step": 6104
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.499231516993623e-05,
+      "loss": 14.6574,
+      "step": 6105
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.499035505548604e-05,
+      "loss": 16.6106,
+      "step": 6106
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.498839457772673e-05,
+      "loss": 17.0101,
+      "step": 6107
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.498643373667409e-05,
+      "loss": 19.3684,
+      "step": 6108
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.4984472532344e-05,
+      "loss": 17.8531,
+      "step": 6109
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.498251096475227e-05,
+      "loss": 17.217,
+      "step": 6110
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.498054903391474e-05,
+      "loss": 14.9076,
+      "step": 6111
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.497858673984727e-05,
+      "loss": 16.0108,
+      "step": 6112
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.497662408256572e-05,
+      "loss": 16.1921,
+      "step": 6113
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.497466106208588e-05,
+      "loss": 15.434,
+      "step": 6114
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.497269767842364e-05,
+      "loss": 15.757,
+      "step": 6115
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.497073393159488e-05,
+      "loss": 15.4479,
+      "step": 6116
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.496876982161542e-05,
+      "loss": 17.69,
+      "step": 6117
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.496680534850113e-05,
+      "loss": 15.8363,
+      "step": 6118
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.496484051226787e-05,
+      "loss": 17.8265,
+      "step": 6119
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.496287531293151e-05,
+      "loss": 14.8331,
+      "step": 6120
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.496090975050793e-05,
+      "loss": 16.6744,
+      "step": 6121
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.4958943825013e-05,
+      "loss": 17.3786,
+      "step": 6122
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.495697753646257e-05,
+      "loss": 16.8236,
+      "step": 6123
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.495501088487255e-05,
+      "loss": 15.4558,
+      "step": 6124
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.495304387025881e-05,
+      "loss": 17.4512,
+      "step": 6125
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.495107649263724e-05,
+      "loss": 17.4493,
+      "step": 6126
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.49491087520237e-05,
+      "loss": 15.5748,
+      "step": 6127
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.494714064843411e-05,
+      "loss": 16.3522,
+      "step": 6128
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.494517218188437e-05,
+      "loss": 18.8239,
+      "step": 6129
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.494320335239034e-05,
+      "loss": 18.9124,
+      "step": 6130
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.494123415996793e-05,
+      "loss": 17.0141,
+      "step": 6131
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.493926460463307e-05,
+      "loss": 16.655,
+      "step": 6132
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.493729468640163e-05,
+      "loss": 16.3081,
+      "step": 6133
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.493532440528952e-05,
+      "loss": 16.4857,
+      "step": 6134
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.493335376131268e-05,
+      "loss": 15.7346,
+      "step": 6135
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.4931382754487e-05,
+      "loss": 18.1219,
+      "step": 6136
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.492941138482841e-05,
+      "loss": 16.4455,
+      "step": 6137
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.492743965235282e-05,
+      "loss": 17.6184,
+      "step": 6138
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.492546755707615e-05,
+      "loss": 16.5945,
+      "step": 6139
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.492349509901433e-05,
+      "loss": 16.2444,
+      "step": 6140
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.492152227818329e-05,
+      "loss": 16.149,
+      "step": 6141
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.491954909459895e-05,
+      "loss": 16.1462,
+      "step": 6142
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.491757554827725e-05,
+      "loss": 16.5773,
+      "step": 6143
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.491560163923414e-05,
+      "loss": 16.0648,
+      "step": 6144
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.491362736748555e-05,
+      "loss": 15.599,
+      "step": 6145
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.491165273304741e-05,
+      "loss": 13.9992,
+      "step": 6146
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.490967773593569e-05,
+      "loss": 16.1727,
+      "step": 6147
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.490770237616631e-05,
+      "loss": 16.2582,
+      "step": 6148
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.490572665375526e-05,
+      "loss": 15.9498,
+      "step": 6149
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.490375056871845e-05,
+      "loss": 16.4098,
+      "step": 6150
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.490177412107187e-05,
+      "loss": 15.9687,
+      "step": 6151
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.489979731083147e-05,
+      "loss": 16.7863,
+      "step": 6152
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.489782013801322e-05,
+      "loss": 17.2194,
+      "step": 6153
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.489584260263307e-05,
+      "loss": 16.1204,
+      "step": 6154
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.4893864704707e-05,
+      "loss": 17.6852,
+      "step": 6155
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.489188644425098e-05,
+      "loss": 17.0672,
+      "step": 6156
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.488990782128098e-05,
+      "loss": 17.4662,
+      "step": 6157
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.488792883581299e-05,
+      "loss": 15.0388,
+      "step": 6158
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.488594948786298e-05,
+      "loss": 14.8992,
+      "step": 6159
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.488396977744694e-05,
+      "loss": 17.0861,
+      "step": 6160
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.488198970458086e-05,
+      "loss": 15.2628,
+      "step": 6161
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.488000926928071e-05,
+      "loss": 15.7932,
+      "step": 6162
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.48780284715625e-05,
+      "loss": 17.0202,
+      "step": 6163
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.487604731144224e-05,
+      "loss": 17.1604,
+      "step": 6164
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.487406578893589e-05,
+      "loss": 15.961,
+      "step": 6165
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.487208390405949e-05,
+      "loss": 16.3255,
+      "step": 6166
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.487010165682901e-05,
+      "loss": 17.8934,
+      "step": 6167
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.48681190472605e-05,
+      "loss": 16.4341,
+      "step": 6168
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.486613607536992e-05,
+      "loss": 17.3593,
+      "step": 6169
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.486415274117331e-05,
+      "loss": 14.9592,
+      "step": 6170
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.486216904468668e-05,
+      "loss": 17.8146,
+      "step": 6171
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.486018498592606e-05,
+      "loss": 14.7611,
+      "step": 6172
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.485820056490746e-05,
+      "loss": 14.6787,
+      "step": 6173
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.485621578164689e-05,
+      "loss": 15.5265,
+      "step": 6174
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.485423063616042e-05,
+      "loss": 16.0683,
+      "step": 6175
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.485224512846404e-05,
+      "loss": 14.9603,
+      "step": 6176
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.485025925857383e-05,
+      "loss": 18.4817,
+      "step": 6177
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.484827302650577e-05,
+      "loss": 15.0056,
+      "step": 6178
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.484628643227592e-05,
+      "loss": 15.1199,
+      "step": 6179
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.484429947590034e-05,
+      "loss": 16.3313,
+      "step": 6180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.484231215739505e-05,
+      "loss": 17.7491,
+      "step": 6181
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.484032447677613e-05,
+      "loss": 15.2279,
+      "step": 6182
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.483833643405961e-05,
+      "loss": 18.4029,
+      "step": 6183
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.483634802926154e-05,
+      "loss": 16.9253,
+      "step": 6184
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.483435926239798e-05,
+      "loss": 15.9336,
+      "step": 6185
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.483237013348498e-05,
+      "loss": 16.1731,
+      "step": 6186
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.483038064253865e-05,
+      "loss": 15.7233,
+      "step": 6187
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.4828390789575e-05,
+      "loss": 16.3146,
+      "step": 6188
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.482640057461012e-05,
+      "loss": 16.2864,
+      "step": 6189
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.482440999766008e-05,
+      "loss": 15.3484,
+      "step": 6190
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.482241905874095e-05,
+      "loss": 18.3026,
+      "step": 6191
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.482042775786883e-05,
+      "loss": 15.7085,
+      "step": 6192
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.481843609505978e-05,
+      "loss": 17.4087,
+      "step": 6193
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.481644407032987e-05,
+      "loss": 14.5469,
+      "step": 6194
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.481445168369521e-05,
+      "loss": 16.4935,
+      "step": 6195
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.48124589351719e-05,
+      "loss": 16.1995,
+      "step": 6196
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.481046582477599e-05,
+      "loss": 15.2613,
+      "step": 6197
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.480847235252361e-05,
+      "loss": 19.4742,
+      "step": 6198
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.480647851843085e-05,
+      "loss": 17.9403,
+      "step": 6199
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.480448432251381e-05,
+      "loss": 18.0355,
+      "step": 6200
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.480248976478857e-05,
+      "loss": 16.3808,
+      "step": 6201
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.480049484527127e-05,
+      "loss": 14.9819,
+      "step": 6202
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.479849956397802e-05,
+      "loss": 15.6824,
+      "step": 6203
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.479650392092489e-05,
+      "loss": 16.3403,
+      "step": 6204
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.479450791612805e-05,
+      "loss": 15.514,
+      "step": 6205
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.479251154960359e-05,
+      "loss": 16.3541,
+      "step": 6206
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.479051482136763e-05,
+      "loss": 16.3392,
+      "step": 6207
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.478851773143628e-05,
+      "loss": 16.3705,
+      "step": 6208
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.478652027982572e-05,
+      "loss": 17.1452,
+      "step": 6209
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.478452246655203e-05,
+      "loss": 17.4228,
+      "step": 6210
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.478252429163135e-05,
+      "loss": 17.2643,
+      "step": 6211
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.478052575507982e-05,
+      "loss": 17.8978,
+      "step": 6212
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.477852685691361e-05,
+      "loss": 15.9151,
+      "step": 6213
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.47765275971488e-05,
+      "loss": 18.2311,
+      "step": 6214
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.477452797580161e-05,
+      "loss": 15.2122,
+      "step": 6215
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.477252799288812e-05,
+      "loss": 14.7897,
+      "step": 6216
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.477052764842451e-05,
+      "loss": 16.2034,
+      "step": 6217
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.476852694242693e-05,
+      "loss": 14.4631,
+      "step": 6218
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.476652587491154e-05,
+      "loss": 16.3057,
+      "step": 6219
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.47645244458945e-05,
+      "loss": 20.5214,
+      "step": 6220
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.476252265539195e-05,
+      "loss": 16.9562,
+      "step": 6221
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.476052050342008e-05,
+      "loss": 17.6734,
+      "step": 6222
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.475851798999506e-05,
+      "loss": 17.3588,
+      "step": 6223
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.475651511513304e-05,
+      "loss": 18.8966,
+      "step": 6224
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.47545118788502e-05,
+      "loss": 17.0247,
+      "step": 6225
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.475250828116273e-05,
+      "loss": 17.3084,
+      "step": 6226
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.47505043220868e-05,
+      "loss": 17.1886,
+      "step": 6227
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.474850000163861e-05,
+      "loss": 17.347,
+      "step": 6228
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.474649531983432e-05,
+      "loss": 17.7144,
+      "step": 6229
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.474449027669012e-05,
+      "loss": 17.9152,
+      "step": 6230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.474248487222222e-05,
+      "loss": 17.4872,
+      "step": 6231
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.47404791064468e-05,
+      "loss": 16.4738,
+      "step": 6232
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.473847297938004e-05,
+      "loss": 17.3678,
+      "step": 6233
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.473646649103818e-05,
+      "loss": 17.5379,
+      "step": 6234
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.473445964143738e-05,
+      "loss": 16.5901,
+      "step": 6235
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.47324524305939e-05,
+      "loss": 17.675,
+      "step": 6236
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.473044485852392e-05,
+      "loss": 15.0583,
+      "step": 6237
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.472843692524363e-05,
+      "loss": 16.0513,
+      "step": 6238
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.472642863076927e-05,
+      "loss": 16.9193,
+      "step": 6239
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.472441997511706e-05,
+      "loss": 16.5668,
+      "step": 6240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.472241095830319e-05,
+      "loss": 14.9388,
+      "step": 6241
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.472040158034392e-05,
+      "loss": 15.6864,
+      "step": 6242
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.471839184125546e-05,
+      "loss": 15.1145,
+      "step": 6243
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.471638174105404e-05,
+      "loss": 16.69,
+      "step": 6244
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.471437127975588e-05,
+      "loss": 18.2888,
+      "step": 6245
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.471236045737724e-05,
+      "loss": 15.6111,
+      "step": 6246
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.471034927393435e-05,
+      "loss": 16.961,
+      "step": 6247
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.470833772944344e-05,
+      "loss": 17.3968,
+      "step": 6248
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.470632582392077e-05,
+      "loss": 16.3928,
+      "step": 6249
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.470431355738257e-05,
+      "loss": 16.7241,
+      "step": 6250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.470230092984509e-05,
+      "loss": 18.0605,
+      "step": 6251
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.470028794132459e-05,
+      "loss": 16.0749,
+      "step": 6252
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.469827459183733e-05,
+      "loss": 16.7608,
+      "step": 6253
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.469626088139956e-05,
+      "loss": 15.1089,
+      "step": 6254
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.469424681002756e-05,
+      "loss": 16.5475,
+      "step": 6255
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.469223237773757e-05,
+      "loss": 16.0873,
+      "step": 6256
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.469021758454586e-05,
+      "loss": 16.7564,
+      "step": 6257
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.46882024304687e-05,
+      "loss": 16.9176,
+      "step": 6258
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.468618691552237e-05,
+      "loss": 18.8689,
+      "step": 6259
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.468417103972316e-05,
+      "loss": 16.7559,
+      "step": 6260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.468215480308731e-05,
+      "loss": 15.9889,
+      "step": 6261
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.468013820563114e-05,
+      "loss": 16.7953,
+      "step": 6262
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.467812124737093e-05,
+      "loss": 17.1889,
+      "step": 6263
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.467610392832294e-05,
+      "loss": 15.8648,
+      "step": 6264
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.467408624850347e-05,
+      "loss": 16.4272,
+      "step": 6265
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.467206820792882e-05,
+      "loss": 16.4968,
+      "step": 6266
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.46700498066153e-05,
+      "loss": 16.6821,
+      "step": 6267
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.466803104457919e-05,
+      "loss": 17.6154,
+      "step": 6268
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.46660119218368e-05,
+      "loss": 17.5556,
+      "step": 6269
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.466399243840443e-05,
+      "loss": 15.9387,
+      "step": 6270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.466197259429839e-05,
+      "loss": 17.1686,
+      "step": 6271
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.465995238953498e-05,
+      "loss": 17.5871,
+      "step": 6272
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.465793182413054e-05,
+      "loss": 18.4977,
+      "step": 6273
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.465591089810138e-05,
+      "loss": 17.2964,
+      "step": 6274
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.46538896114638e-05,
+      "loss": 17.5311,
+      "step": 6275
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.465186796423414e-05,
+      "loss": 15.848,
+      "step": 6276
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.46498459564287e-05,
+      "loss": 16.2913,
+      "step": 6277
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.464782358806383e-05,
+      "loss": 15.8628,
+      "step": 6278
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.464580085915587e-05,
+      "loss": 16.2977,
+      "step": 6279
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.464377776972114e-05,
+      "loss": 15.9918,
+      "step": 6280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.464175431977597e-05,
+      "loss": 14.3594,
+      "step": 6281
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.463973050933674e-05,
+      "loss": 15.9542,
+      "step": 6282
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.463770633841973e-05,
+      "loss": 15.0856,
+      "step": 6283
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.463568180704132e-05,
+      "loss": 20.1487,
+      "step": 6284
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.463365691521787e-05,
+      "loss": 15.8866,
+      "step": 6285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.463163166296572e-05,
+      "loss": 16.366,
+      "step": 6286
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.46296060503012e-05,
+      "loss": 15.1959,
+      "step": 6287
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.46275800772407e-05,
+      "loss": 17.9024,
+      "step": 6288
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.462555374380058e-05,
+      "loss": 16.5195,
+      "step": 6289
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.46235270499972e-05,
+      "loss": 16.7346,
+      "step": 6290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.462149999584689e-05,
+      "loss": 15.8194,
+      "step": 6291
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.461947258136607e-05,
+      "loss": 17.1197,
+      "step": 6292
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.461744480657108e-05,
+      "loss": 15.0248,
+      "step": 6293
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.461541667147831e-05,
+      "loss": 16.9946,
+      "step": 6294
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.461338817610413e-05,
+      "loss": 15.6112,
+      "step": 6295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.461135932046494e-05,
+      "loss": 18.0191,
+      "step": 6296
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.460933010457708e-05,
+      "loss": 16.4623,
+      "step": 6297
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.460730052845698e-05,
+      "loss": 18.533,
+      "step": 6298
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.460527059212101e-05,
+      "loss": 17.0732,
+      "step": 6299
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.460324029558556e-05,
+      "loss": 19.5915,
+      "step": 6300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.460120963886703e-05,
+      "loss": 15.8967,
+      "step": 6301
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.459917862198182e-05,
+      "loss": 17.8644,
+      "step": 6302
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.459714724494634e-05,
+      "loss": 15.6906,
+      "step": 6303
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.459511550777698e-05,
+      "loss": 16.7152,
+      "step": 6304
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.459308341049014e-05,
+      "loss": 16.785,
+      "step": 6305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.459105095310223e-05,
+      "loss": 15.4602,
+      "step": 6306
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.45890181356297e-05,
+      "loss": 17.3841,
+      "step": 6307
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.458698495808892e-05,
+      "loss": 18.138,
+      "step": 6308
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.458495142049633e-05,
+      "loss": 15.5178,
+      "step": 6309
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.458291752286835e-05,
+      "loss": 17.6537,
+      "step": 6310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45808832652214e-05,
+      "loss": 16.991,
+      "step": 6311
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.457884864757192e-05,
+      "loss": 17.3747,
+      "step": 6312
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.457681366993632e-05,
+      "loss": 19.2474,
+      "step": 6313
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.457477833233103e-05,
+      "loss": 16.981,
+      "step": 6314
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.457274263477251e-05,
+      "loss": 15.582,
+      "step": 6315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.457070657727718e-05,
+      "loss": 17.5778,
+      "step": 6316
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45686701598615e-05,
+      "loss": 15.8996,
+      "step": 6317
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45666333825419e-05,
+      "loss": 15.2892,
+      "step": 6318
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.456459624533483e-05,
+      "loss": 16.3491,
+      "step": 6319
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.456255874825675e-05,
+      "loss": 17.5916,
+      "step": 6320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.456052089132408e-05,
+      "loss": 15.4422,
+      "step": 6321
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.455848267455332e-05,
+      "loss": 16.8653,
+      "step": 6322
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45564440979609e-05,
+      "loss": 16.451,
+      "step": 6323
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45544051615633e-05,
+      "loss": 14.9397,
+      "step": 6324
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.455236586537697e-05,
+      "loss": 15.8637,
+      "step": 6325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45503262094184e-05,
+      "loss": 15.9562,
+      "step": 6326
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.454828619370403e-05,
+      "loss": 15.5778,
+      "step": 6327
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.454624581825036e-05,
+      "loss": 16.2973,
+      "step": 6328
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.454420508307385e-05,
+      "loss": 17.1692,
+      "step": 6329
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.454216398819098e-05,
+      "loss": 16.4111,
+      "step": 6330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.454012253361825e-05,
+      "loss": 18.1418,
+      "step": 6331
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.453808071937213e-05,
+      "loss": 17.9418,
+      "step": 6332
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45360385454691e-05,
+      "loss": 15.4145,
+      "step": 6333
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.453399601192567e-05,
+      "loss": 17.6718,
+      "step": 6334
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.453195311875834e-05,
+      "loss": 15.5492,
+      "step": 6335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.452990986598356e-05,
+      "loss": 17.7724,
+      "step": 6336
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45278662536179e-05,
+      "loss": 16.9276,
+      "step": 6337
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.452582228167779e-05,
+      "loss": 17.754,
+      "step": 6338
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45237779501798e-05,
+      "loss": 17.3499,
+      "step": 6339
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.452173325914039e-05,
+      "loss": 14.581,
+      "step": 6340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.451968820857611e-05,
+      "loss": 15.7863,
+      "step": 6341
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.451764279850343e-05,
+      "loss": 17.3741,
+      "step": 6342
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.451559702893891e-05,
+      "loss": 17.2214,
+      "step": 6343
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.451355089989906e-05,
+      "loss": 15.5807,
+      "step": 6344
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.451150441140039e-05,
+      "loss": 17.3574,
+      "step": 6345
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.450945756345942e-05,
+      "loss": 16.4766,
+      "step": 6346
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45074103560927e-05,
+      "loss": 16.3636,
+      "step": 6347
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.450536278931677e-05,
+      "loss": 15.8485,
+      "step": 6348
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.450331486314811e-05,
+      "loss": 18.9465,
+      "step": 6349
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.450126657760332e-05,
+      "loss": 16.5309,
+      "step": 6350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.449921793269891e-05,
+      "loss": 14.6855,
+      "step": 6351
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.449716892845144e-05,
+      "loss": 18.7602,
+      "step": 6352
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.449511956487742e-05,
+      "loss": 15.5967,
+      "step": 6353
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.449306984199344e-05,
+      "loss": 16.1309,
+      "step": 6354
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.449101975981604e-05,
+      "loss": 16.5171,
+      "step": 6355
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.448896931836176e-05,
+      "loss": 19.293,
+      "step": 6356
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.448691851764717e-05,
+      "loss": 15.7275,
+      "step": 6357
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.448486735768884e-05,
+      "loss": 16.547,
+      "step": 6358
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.448281583850332e-05,
+      "loss": 18.0276,
+      "step": 6359
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.448076396010716e-05,
+      "loss": 16.8746,
+      "step": 6360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.447871172251698e-05,
+      "loss": 16.134,
+      "step": 6361
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44766591257493e-05,
+      "loss": 17.7409,
+      "step": 6362
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.447460616982074e-05,
+      "loss": 15.9834,
+      "step": 6363
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.447255285474783e-05,
+      "loss": 17.4788,
+      "step": 6364
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.447049918054718e-05,
+      "loss": 16.6169,
+      "step": 6365
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.446844514723539e-05,
+      "loss": 15.2153,
+      "step": 6366
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.4466390754829e-05,
+      "loss": 16.6391,
+      "step": 6367
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.446433600334465e-05,
+      "loss": 16.7468,
+      "step": 6368
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44622808927989e-05,
+      "loss": 15.8165,
+      "step": 6369
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.446022542320834e-05,
+      "loss": 16.8907,
+      "step": 6370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44581695945896e-05,
+      "loss": 15.3947,
+      "step": 6371
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.445611340695926e-05,
+      "loss": 16.4702,
+      "step": 6372
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.445405686033393e-05,
+      "loss": 18.1558,
+      "step": 6373
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.445199995473022e-05,
+      "loss": 16.3694,
+      "step": 6374
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.444994269016474e-05,
+      "loss": 17.3913,
+      "step": 6375
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.444788506665409e-05,
+      "loss": 16.7788,
+      "step": 6376
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44458270842149e-05,
+      "loss": 15.0112,
+      "step": 6377
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.444376874286378e-05,
+      "loss": 16.2294,
+      "step": 6378
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.444171004261735e-05,
+      "loss": 16.5497,
+      "step": 6379
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.443965098349226e-05,
+      "loss": 15.9855,
+      "step": 6380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.443759156550511e-05,
+      "loss": 16.8652,
+      "step": 6381
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.443553178867253e-05,
+      "loss": 19.3726,
+      "step": 6382
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.443347165301116e-05,
+      "loss": 15.8464,
+      "step": 6383
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.443141115853764e-05,
+      "loss": 16.6449,
+      "step": 6384
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.442935030526861e-05,
+      "loss": 17.2596,
+      "step": 6385
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.442728909322071e-05,
+      "loss": 15.782,
+      "step": 6386
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.442522752241058e-05,
+      "loss": 15.5777,
+      "step": 6387
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.442316559285487e-05,
+      "loss": 15.3563,
+      "step": 6388
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.442110330457023e-05,
+      "loss": 17.3006,
+      "step": 6389
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.441904065757331e-05,
+      "loss": 15.1787,
+      "step": 6390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44169776518808e-05,
+      "loss": 15.6017,
+      "step": 6391
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44149142875093e-05,
+      "loss": 16.0195,
+      "step": 6392
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44128505644755e-05,
+      "loss": 14.0479,
+      "step": 6393
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.441078648279608e-05,
+      "loss": 16.4099,
+      "step": 6394
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.440872204248772e-05,
+      "loss": 18.4142,
+      "step": 6395
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.440665724356703e-05,
+      "loss": 16.8839,
+      "step": 6396
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.440459208605075e-05,
+      "loss": 15.5026,
+      "step": 6397
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.440252656995551e-05,
+      "loss": 15.9969,
+      "step": 6398
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.4400460695298e-05,
+      "loss": 17.4698,
+      "step": 6399
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.439839446209493e-05,
+      "loss": 15.9666,
+      "step": 6400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.439632787036295e-05,
+      "loss": 15.7254,
+      "step": 6401
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.439426092011875e-05,
+      "loss": 16.2355,
+      "step": 6402
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.439219361137907e-05,
+      "loss": 17.2516,
+      "step": 6403
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.439012594416055e-05,
+      "loss": 17.4546,
+      "step": 6404
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.438805791847991e-05,
+      "loss": 16.2335,
+      "step": 6405
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.438598953435384e-05,
+      "loss": 16.3156,
+      "step": 6406
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.438392079179906e-05,
+      "loss": 16.18,
+      "step": 6407
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.438185169083224e-05,
+      "loss": 14.6299,
+      "step": 6408
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.437978223147012e-05,
+      "loss": 14.7304,
+      "step": 6409
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.437771241372942e-05,
+      "loss": 14.7985,
+      "step": 6410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.437564223762683e-05,
+      "loss": 17.3244,
+      "step": 6411
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.437357170317906e-05,
+      "loss": 15.0357,
+      "step": 6412
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.437150081040287e-05,
+      "loss": 20.649,
+      "step": 6413
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.436942955931493e-05,
+      "loss": 16.6538,
+      "step": 6414
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.436735794993202e-05,
+      "loss": 15.0278,
+      "step": 6415
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.436528598227083e-05,
+      "loss": 17.1148,
+      "step": 6416
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.436321365634811e-05,
+      "loss": 18.0847,
+      "step": 6417
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.43611409721806e-05,
+      "loss": 15.7051,
+      "step": 6418
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.435906792978502e-05,
+      "loss": 16.2487,
+      "step": 6419
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.43569945291781e-05,
+      "loss": 17.413,
+      "step": 6420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.435492077037662e-05,
+      "loss": 15.3347,
+      "step": 6421
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.43528466533973e-05,
+      "loss": 16.0135,
+      "step": 6422
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.43507721782569e-05,
+      "loss": 18.1753,
+      "step": 6423
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.434869734497216e-05,
+      "loss": 16.8483,
+      "step": 6424
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.434662215355985e-05,
+      "loss": 15.0431,
+      "step": 6425
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.434454660403671e-05,
+      "loss": 16.2689,
+      "step": 6426
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.434247069641952e-05,
+      "loss": 17.1335,
+      "step": 6427
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.434039443072501e-05,
+      "loss": 17.8341,
+      "step": 6428
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.433831780697001e-05,
+      "loss": 15.5229,
+      "step": 6429
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.433624082517123e-05,
+      "loss": 15.3454,
+      "step": 6430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.433416348534547e-05,
+      "loss": 17.7535,
+      "step": 6431
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.433208578750949e-05,
+      "loss": 15.9248,
+      "step": 6432
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.433000773168008e-05,
+      "loss": 17.3528,
+      "step": 6433
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.4327929317874e-05,
+      "loss": 17.126,
+      "step": 6434
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.432585054610807e-05,
+      "loss": 17.1721,
+      "step": 6435
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.432377141639906e-05,
+      "loss": 16.83,
+      "step": 6436
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.432169192876375e-05,
+      "loss": 16.5774,
+      "step": 6437
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.431961208321892e-05,
+      "loss": 16.3683,
+      "step": 6438
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.43175318797814e-05,
+      "loss": 16.2024,
+      "step": 6439
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.431545131846797e-05,
+      "loss": 16.8366,
+      "step": 6440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.431337039929543e-05,
+      "loss": 16.6831,
+      "step": 6441
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.43112891222806e-05,
+      "loss": 16.875,
+      "step": 6442
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.430920748744026e-05,
+      "loss": 17.0824,
+      "step": 6443
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.430712549479124e-05,
+      "loss": 15.9587,
+      "step": 6444
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.430504314435034e-05,
+      "loss": 17.5304,
+      "step": 6445
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.430296043613438e-05,
+      "loss": 16.9558,
+      "step": 6446
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.43008773701602e-05,
+      "loss": 16.9646,
+      "step": 6447
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.429879394644458e-05,
+      "loss": 17.3248,
+      "step": 6448
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.429671016500436e-05,
+      "loss": 17.9503,
+      "step": 6449
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.429462602585639e-05,
+      "loss": 16.4096,
+      "step": 6450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.429254152901748e-05,
+      "loss": 17.056,
+      "step": 6451
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.429045667450447e-05,
+      "loss": 18.0589,
+      "step": 6452
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.428837146233419e-05,
+      "loss": 16.87,
+      "step": 6453
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.428628589252348e-05,
+      "loss": 15.8538,
+      "step": 6454
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.428419996508917e-05,
+      "loss": 16.095,
+      "step": 6455
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.428211368004814e-05,
+      "loss": 16.1911,
+      "step": 6456
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.428002703741718e-05,
+      "loss": 17.6819,
+      "step": 6457
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.42779400372132e-05,
+      "loss": 16.0857,
+      "step": 6458
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.427585267945302e-05,
+      "loss": 17.576,
+      "step": 6459
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.42737649641535e-05,
+      "loss": 17.5039,
+      "step": 6460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.42716768913315e-05,
+      "loss": 14.64,
+      "step": 6461
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.426958846100389e-05,
+      "loss": 16.9325,
+      "step": 6462
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.426749967318751e-05,
+      "loss": 17.0572,
+      "step": 6463
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.426541052789925e-05,
+      "loss": 18.1583,
+      "step": 6464
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.4263321025156e-05,
+      "loss": 15.5256,
+      "step": 6465
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.426123116497458e-05,
+      "loss": 17.1721,
+      "step": 6466
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.425914094737192e-05,
+      "loss": 15.4077,
+      "step": 6467
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.425705037236486e-05,
+      "loss": 18.0296,
+      "step": 6468
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.42549594399703e-05,
+      "loss": 16.2951,
+      "step": 6469
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.425286815020513e-05,
+      "loss": 17.0994,
+      "step": 6470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.425077650308621e-05,
+      "loss": 16.3759,
+      "step": 6471
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.424868449863048e-05,
+      "loss": 15.0265,
+      "step": 6472
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.424659213685476e-05,
+      "loss": 15.5116,
+      "step": 6473
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.424449941777602e-05,
+      "loss": 17.6308,
+      "step": 6474
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.424240634141112e-05,
+      "loss": 17.0254,
+      "step": 6475
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.424031290777698e-05,
+      "loss": 16.1511,
+      "step": 6476
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.423821911689049e-05,
+      "loss": 15.8143,
+      "step": 6477
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.423612496876855e-05,
+      "loss": 18.2444,
+      "step": 6478
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.423403046342809e-05,
+      "loss": 15.2107,
+      "step": 6479
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.423193560088602e-05,
+      "loss": 15.8815,
+      "step": 6480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.422984038115926e-05,
+      "loss": 16.1445,
+      "step": 6481
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.422774480426472e-05,
+      "loss": 16.7436,
+      "step": 6482
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.422564887021933e-05,
+      "loss": 15.9024,
+      "step": 6483
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.422355257904e-05,
+      "loss": 19.1075,
+      "step": 6484
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.42214559307437e-05,
+      "loss": 17.2948,
+      "step": 6485
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.42193589253473e-05,
+      "loss": 16.1735,
+      "step": 6486
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.421726156286777e-05,
+      "loss": 17.5368,
+      "step": 6487
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.421516384332205e-05,
+      "loss": 15.8655,
+      "step": 6488
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.421306576672707e-05,
+      "loss": 17.4764,
+      "step": 6489
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.421096733309978e-05,
+      "loss": 16.7182,
+      "step": 6490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.420886854245712e-05,
+      "loss": 16.3227,
+      "step": 6491
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.420676939481602e-05,
+      "loss": 16.9962,
+      "step": 6492
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.420466989019345e-05,
+      "loss": 14.8581,
+      "step": 6493
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.420257002860638e-05,
+      "loss": 13.8405,
+      "step": 6494
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.420046981007173e-05,
+      "loss": 18.4761,
+      "step": 6495
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.41983692346065e-05,
+      "loss": 17.4224,
+      "step": 6496
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.419626830222762e-05,
+      "loss": 14.8775,
+      "step": 6497
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.419416701295208e-05,
+      "loss": 18.8958,
+      "step": 6498
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.419206536679681e-05,
+      "loss": 17.254,
+      "step": 6499
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.418996336377883e-05,
+      "loss": 16.9542,
+      "step": 6500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.418786100391508e-05,
+      "loss": 18.1902,
+      "step": 6501
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.418575828722256e-05,
+      "loss": 16.4148,
+      "step": 6502
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.418365521371823e-05,
+      "loss": 17.9111,
+      "step": 6503
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.418155178341907e-05,
+      "loss": 18.2503,
+      "step": 6504
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.417944799634211e-05,
+      "loss": 17.676,
+      "step": 6505
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.417734385250428e-05,
+      "loss": 16.8855,
+      "step": 6506
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.417523935192261e-05,
+      "loss": 16.1095,
+      "step": 6507
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.417313449461408e-05,
+      "loss": 17.5317,
+      "step": 6508
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.417102928059569e-05,
+      "loss": 16.8667,
+      "step": 6509
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.416892370988444e-05,
+      "loss": 16.2135,
+      "step": 6510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.416681778249732e-05,
+      "loss": 16.2945,
+      "step": 6511
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.416471149845136e-05,
+      "loss": 17.5326,
+      "step": 6512
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.416260485776356e-05,
+      "loss": 16.852,
+      "step": 6513
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.416049786045092e-05,
+      "loss": 16.6328,
+      "step": 6514
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.415839050653048e-05,
+      "loss": 18.2871,
+      "step": 6515
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.415628279601923e-05,
+      "loss": 16.721,
+      "step": 6516
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.41541747289342e-05,
+      "loss": 17.2416,
+      "step": 6517
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.415206630529241e-05,
+      "loss": 15.1051,
+      "step": 6518
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.41499575251109e-05,
+      "loss": 17.6101,
+      "step": 6519
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.414784838840669e-05,
+      "loss": 17.4079,
+      "step": 6520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.41457388951968e-05,
+      "loss": 17.8233,
+      "step": 6521
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.414362904549828e-05,
+      "loss": 16.43,
+      "step": 6522
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.414151883932818e-05,
+      "loss": 15.4804,
+      "step": 6523
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.413940827670352e-05,
+      "loss": 16.3884,
+      "step": 6524
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.413729735764131e-05,
+      "loss": 16.7492,
+      "step": 6525
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.413518608215868e-05,
+      "loss": 16.8615,
+      "step": 6526
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.413307445027262e-05,
+      "loss": 16.3137,
+      "step": 6527
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.413096246200019e-05,
+      "loss": 15.0479,
+      "step": 6528
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.412885011735844e-05,
+      "loss": 18.1783,
+      "step": 6529
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.412673741636445e-05,
+      "loss": 14.4834,
+      "step": 6530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.412462435903528e-05,
+      "loss": 15.3726,
+      "step": 6531
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.412251094538796e-05,
+      "loss": 16.4426,
+      "step": 6532
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.412039717543959e-05,
+      "loss": 17.919,
+      "step": 6533
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.411828304920722e-05,
+      "loss": 16.7494,
+      "step": 6534
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.411616856670795e-05,
+      "loss": 17.7696,
+      "step": 6535
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.411405372795881e-05,
+      "loss": 16.5358,
+      "step": 6536
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.41119385329769e-05,
+      "loss": 16.745,
+      "step": 6537
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.410982298177933e-05,
+      "loss": 18.3201,
+      "step": 6538
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.410770707438313e-05,
+      "loss": 17.1455,
+      "step": 6539
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.410559081080544e-05,
+      "loss": 17.7376,
+      "step": 6540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.410347419106329e-05,
+      "loss": 15.7472,
+      "step": 6541
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.410135721517382e-05,
+      "loss": 17.1172,
+      "step": 6542
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.409923988315412e-05,
+      "loss": 17.2431,
+      "step": 6543
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.409712219502128e-05,
+      "loss": 16.79,
+      "step": 6544
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.409500415079238e-05,
+      "loss": 15.7342,
+      "step": 6545
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.409288575048455e-05,
+      "loss": 15.5832,
+      "step": 6546
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.409076699411488e-05,
+      "loss": 15.0623,
+      "step": 6547
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.40886478817005e-05,
+      "loss": 18.9912,
+      "step": 6548
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.408652841325851e-05,
+      "loss": 17.125,
+      "step": 6549
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.408440858880603e-05,
+      "loss": 17.8015,
+      "step": 6550
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.408228840836018e-05,
+      "loss": 15.4633,
+      "step": 6551
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.408016787193808e-05,
+      "loss": 17.7994,
+      "step": 6552
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.407804697955682e-05,
+      "loss": 17.6117,
+      "step": 6553
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.407592573123358e-05,
+      "loss": 17.4515,
+      "step": 6554
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.407380412698546e-05,
+      "loss": 16.7425,
+      "step": 6555
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.407168216682962e-05,
+      "loss": 17.5346,
+      "step": 6556
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.406955985078313e-05,
+      "loss": 17.0343,
+      "step": 6557
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.406743717886321e-05,
+      "loss": 16.6948,
+      "step": 6558
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.406531415108694e-05,
+      "loss": 14.8868,
+      "step": 6559
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.40631907674715e-05,
+      "loss": 16.8789,
+      "step": 6560
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.406106702803404e-05,
+      "loss": 15.3921,
+      "step": 6561
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.405894293279168e-05,
+      "loss": 15.8279,
+      "step": 6562
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.405681848176157e-05,
+      "loss": 18.0437,
+      "step": 6563
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.405469367496089e-05,
+      "loss": 16.5094,
+      "step": 6564
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.40525685124068e-05,
+      "loss": 17.4075,
+      "step": 6565
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.405044299411646e-05,
+      "loss": 17.5227,
+      "step": 6566
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.4048317120107e-05,
+      "loss": 15.4125,
+      "step": 6567
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.404619089039564e-05,
+      "loss": 15.9172,
+      "step": 6568
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.40440643049995e-05,
+      "loss": 16.3454,
+      "step": 6569
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.404193736393578e-05,
+      "loss": 15.8949,
+      "step": 6570
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.403981006722167e-05,
+      "loss": 16.6297,
+      "step": 6571
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.40376824148743e-05,
+      "loss": 16.2274,
+      "step": 6572
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.403555440691091e-05,
+      "loss": 17.7988,
+      "step": 6573
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.403342604334864e-05,
+      "loss": 15.2632,
+      "step": 6574
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.40312973242047e-05,
+      "loss": 16.3828,
+      "step": 6575
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.402916824949626e-05,
+      "loss": 17.6121,
+      "step": 6576
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.402703881924053e-05,
+      "loss": 15.6911,
+      "step": 6577
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.402490903345472e-05,
+      "loss": 16.4827,
+      "step": 6578
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.402277889215599e-05,
+      "loss": 15.9465,
+      "step": 6579
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.402064839536157e-05,
+      "loss": 17.6593,
+      "step": 6580
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.401851754308865e-05,
+      "loss": 17.7042,
+      "step": 6581
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.401638633535443e-05,
+      "loss": 16.64,
+      "step": 6582
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.401425477217616e-05,
+      "loss": 16.0417,
+      "step": 6583
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.4012122853571e-05,
+      "loss": 15.0377,
+      "step": 6584
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.400999057955621e-05,
+      "loss": 18.6979,
+      "step": 6585
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.400785795014898e-05,
+      "loss": 14.8179,
+      "step": 6586
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.400572496536655e-05,
+      "loss": 16.524,
+      "step": 6587
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.400359162522613e-05,
+      "loss": 17.1968,
+      "step": 6588
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.400145792974496e-05,
+      "loss": 16.164,
+      "step": 6589
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.399932387894025e-05,
+      "loss": 16.6085,
+      "step": 6590
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.399718947282925e-05,
+      "loss": 16.8647,
+      "step": 6591
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.39950547114292e-05,
+      "loss": 17.1426,
+      "step": 6592
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.399291959475733e-05,
+      "loss": 16.3727,
+      "step": 6593
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.399078412283089e-05,
+      "loss": 17.2926,
+      "step": 6594
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.398864829566709e-05,
+      "loss": 16.8019,
+      "step": 6595
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.398651211328323e-05,
+      "loss": 16.3445,
+      "step": 6596
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.398437557569652e-05,
+      "loss": 16.6404,
+      "step": 6597
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.398223868292424e-05,
+      "loss": 17.6836,
+      "step": 6598
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.398010143498364e-05,
+      "loss": 14.9355,
+      "step": 6599
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.397796383189197e-05,
+      "loss": 16.3219,
+      "step": 6600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.397582587366649e-05,
+      "loss": 16.4171,
+      "step": 6601
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.397368756032445e-05,
+      "loss": 15.6389,
+      "step": 6602
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.397154889188317e-05,
+      "loss": 18.2569,
+      "step": 6603
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.396940986835987e-05,
+      "loss": 15.3435,
+      "step": 6604
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.396727048977183e-05,
+      "loss": 16.6172,
+      "step": 6605
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.396513075613636e-05,
+      "loss": 14.6311,
+      "step": 6606
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.39629906674707e-05,
+      "loss": 18.5257,
+      "step": 6607
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.396085022379216e-05,
+      "loss": 18.0634,
+      "step": 6608
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.395870942511799e-05,
+      "loss": 16.8234,
+      "step": 6609
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.395656827146551e-05,
+      "loss": 14.3402,
+      "step": 6610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.3954426762852e-05,
+      "loss": 16.5734,
+      "step": 6611
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.395228489929475e-05,
+      "loss": 16.2794,
+      "step": 6612
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.395014268081106e-05,
+      "loss": 15.8437,
+      "step": 6613
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.394800010741823e-05,
+      "loss": 16.7714,
+      "step": 6614
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.394585717913355e-05,
+      "loss": 15.617,
+      "step": 6615
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.394371389597436e-05,
+      "loss": 16.4999,
+      "step": 6616
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.394157025795792e-05,
+      "loss": 16.6474,
+      "step": 6617
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.393942626510157e-05,
+      "loss": 16.6521,
+      "step": 6618
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.393728191742261e-05,
+      "loss": 15.8672,
+      "step": 6619
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.393513721493837e-05,
+      "loss": 16.6798,
+      "step": 6620
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.393299215766616e-05,
+      "loss": 15.6533,
+      "step": 6621
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.393084674562329e-05,
+      "loss": 16.4256,
+      "step": 6622
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.392870097882711e-05,
+      "loss": 16.6528,
+      "step": 6623
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.392655485729494e-05,
+      "loss": 15.5663,
+      "step": 6624
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.39244083810441e-05,
+      "loss": 15.2455,
+      "step": 6625
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.392226155009193e-05,
+      "loss": 17.4303,
+      "step": 6626
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.392011436445575e-05,
+      "loss": 16.2638,
+      "step": 6627
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.391796682415295e-05,
+      "loss": 17.5303,
+      "step": 6628
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.39158189292008e-05,
+      "loss": 17.1433,
+      "step": 6629
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.39136706796167e-05,
+      "loss": 15.3246,
+      "step": 6630
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.391152207541797e-05,
+      "loss": 15.3581,
+      "step": 6631
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.390937311662197e-05,
+      "loss": 16.4842,
+      "step": 6632
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.390722380324607e-05,
+      "loss": 16.3177,
+      "step": 6633
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.39050741353076e-05,
+      "loss": 16.3291,
+      "step": 6634
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.390292411282393e-05,
+      "loss": 15.9932,
+      "step": 6635
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.39007737358124e-05,
+      "loss": 16.1037,
+      "step": 6636
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.389862300429042e-05,
+      "loss": 17.3577,
+      "step": 6637
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.389647191827533e-05,
+      "loss": 16.1359,
+      "step": 6638
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.38943204777845e-05,
+      "loss": 15.8258,
+      "step": 6639
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.38921686828353e-05,
+      "loss": 15.9083,
+      "step": 6640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.389001653344512e-05,
+      "loss": 16.9064,
+      "step": 6641
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.388786402963132e-05,
+      "loss": 17.0585,
+      "step": 6642
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.388571117141131e-05,
+      "loss": 17.0442,
+      "step": 6643
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.388355795880246e-05,
+      "loss": 16.9675,
+      "step": 6644
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.388140439182216e-05,
+      "loss": 16.2545,
+      "step": 6645
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.387925047048778e-05,
+      "loss": 15.7588,
+      "step": 6646
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.387709619481674e-05,
+      "loss": 16.513,
+      "step": 6647
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.387494156482643e-05,
+      "loss": 16.6194,
+      "step": 6648
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.387278658053424e-05,
+      "loss": 17.2126,
+      "step": 6649
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.38706312419576e-05,
+      "loss": 16.0966,
+      "step": 6650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.386847554911387e-05,
+      "loss": 14.0972,
+      "step": 6651
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.386631950202049e-05,
+      "loss": 16.1713,
+      "step": 6652
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.386416310069488e-05,
+      "loss": 16.1441,
+      "step": 6653
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.386200634515441e-05,
+      "loss": 15.7866,
+      "step": 6654
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.385984923541652e-05,
+      "loss": 14.6853,
+      "step": 6655
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.385769177149865e-05,
+      "loss": 17.9162,
+      "step": 6656
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.385553395341821e-05,
+      "loss": 15.6457,
+      "step": 6657
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.385337578119259e-05,
+      "loss": 17.213,
+      "step": 6658
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.385121725483927e-05,
+      "loss": 16.6603,
+      "step": 6659
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.384905837437565e-05,
+      "loss": 15.3264,
+      "step": 6660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.384689913981918e-05,
+      "loss": 13.9393,
+      "step": 6661
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.384473955118726e-05,
+      "loss": 16.3813,
+      "step": 6662
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.384257960849739e-05,
+      "loss": 16.4984,
+      "step": 6663
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.384041931176697e-05,
+      "loss": 16.3497,
+      "step": 6664
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.383825866101343e-05,
+      "loss": 15.408,
+      "step": 6665
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.383609765625427e-05,
+      "loss": 14.4579,
+      "step": 6666
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.38339362975069e-05,
+      "loss": 14.7061,
+      "step": 6667
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.383177458478878e-05,
+      "loss": 15.0574,
+      "step": 6668
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.382961251811738e-05,
+      "loss": 14.8141,
+      "step": 6669
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.382745009751015e-05,
+      "loss": 15.4357,
+      "step": 6670
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.382528732298454e-05,
+      "loss": 15.3108,
+      "step": 6671
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.382312419455804e-05,
+      "loss": 15.5443,
+      "step": 6672
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.382096071224811e-05,
+      "loss": 15.4058,
+      "step": 6673
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.381879687607221e-05,
+      "loss": 18.6086,
+      "step": 6674
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.381663268604782e-05,
+      "loss": 14.8529,
+      "step": 6675
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.381446814219242e-05,
+      "loss": 17.3085,
+      "step": 6676
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.381230324452347e-05,
+      "loss": 15.3443,
+      "step": 6677
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.38101379930585e-05,
+      "loss": 15.2581,
+      "step": 6678
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.380797238781494e-05,
+      "loss": 15.3042,
+      "step": 6679
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.380580642881031e-05,
+      "loss": 15.485,
+      "step": 6680
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.380364011606208e-05,
+      "loss": 16.4999,
+      "step": 6681
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.380147344958777e-05,
+      "loss": 17.0829,
+      "step": 6682
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.379930642940487e-05,
+      "loss": 16.2376,
+      "step": 6683
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.379713905553085e-05,
+      "loss": 18.1223,
+      "step": 6684
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.379497132798323e-05,
+      "loss": 15.6332,
+      "step": 6685
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.379280324677954e-05,
+      "loss": 16.3299,
+      "step": 6686
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.379063481193725e-05,
+      "loss": 15.3891,
+      "step": 6687
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.378846602347388e-05,
+      "loss": 17.6514,
+      "step": 6688
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.378629688140696e-05,
+      "loss": 15.4806,
+      "step": 6689
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.378412738575401e-05,
+      "loss": 17.4066,
+      "step": 6690
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.378195753653252e-05,
+      "loss": 14.6632,
+      "step": 6691
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.377978733376002e-05,
+      "loss": 15.2997,
+      "step": 6692
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.377761677745407e-05,
+      "loss": 18.2232,
+      "step": 6693
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.377544586763215e-05,
+      "loss": 17.0886,
+      "step": 6694
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.377327460431182e-05,
+      "loss": 16.7476,
+      "step": 6695
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.377110298751059e-05,
+      "loss": 17.288,
+      "step": 6696
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.376893101724602e-05,
+      "loss": 17.4086,
+      "step": 6697
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.376675869353565e-05,
+      "loss": 16.763,
+      "step": 6698
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.376458601639701e-05,
+      "loss": 16.5786,
+      "step": 6699
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.376241298584764e-05,
+      "loss": 16.1885,
+      "step": 6700
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.37602396019051e-05,
+      "loss": 17.0291,
+      "step": 6701
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.375806586458693e-05,
+      "loss": 16.3961,
+      "step": 6702
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.375589177391068e-05,
+      "loss": 15.5347,
+      "step": 6703
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.375371732989393e-05,
+      "loss": 16.3166,
+      "step": 6704
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.375154253255423e-05,
+      "loss": 15.4885,
+      "step": 6705
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.374936738190914e-05,
+      "loss": 16.7953,
+      "step": 6706
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.37471918779762e-05,
+      "loss": 15.3544,
+      "step": 6707
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.374501602077301e-05,
+      "loss": 17.6779,
+      "step": 6708
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.374283981031714e-05,
+      "loss": 15.5918,
+      "step": 6709
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.374066324662615e-05,
+      "loss": 16.6979,
+      "step": 6710
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.373848632971763e-05,
+      "loss": 15.8291,
+      "step": 6711
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.373630905960913e-05,
+      "loss": 17.4033,
+      "step": 6712
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.373413143631826e-05,
+      "loss": 17.04,
+      "step": 6713
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.373195345986259e-05,
+      "loss": 16.845,
+      "step": 6714
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.372977513025971e-05,
+      "loss": 14.0176,
+      "step": 6715
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.372759644752722e-05,
+      "loss": 15.6695,
+      "step": 6716
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.372541741168272e-05,
+      "loss": 15.9813,
+      "step": 6717
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.372323802274379e-05,
+      "loss": 15.1346,
+      "step": 6718
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.372105828072803e-05,
+      "loss": 14.7532,
+      "step": 6719
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.371887818565304e-05,
+      "loss": 17.5311,
+      "step": 6720
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.371669773753644e-05,
+      "loss": 19.8944,
+      "step": 6721
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.371451693639582e-05,
+      "loss": 16.0584,
+      "step": 6722
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.371233578224881e-05,
+      "loss": 15.8308,
+      "step": 6723
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.3710154275113e-05,
+      "loss": 17.454,
+      "step": 6724
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.370797241500602e-05,
+      "loss": 16.1813,
+      "step": 6725
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.37057902019455e-05,
+      "loss": 14.3127,
+      "step": 6726
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.370360763594904e-05,
+      "loss": 16.6257,
+      "step": 6727
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.370142471703426e-05,
+      "loss": 15.6601,
+      "step": 6728
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.369924144521882e-05,
+      "loss": 18.5659,
+      "step": 6729
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.369705782052034e-05,
+      "loss": 15.211,
+      "step": 6730
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.369487384295644e-05,
+      "loss": 15.9383,
+      "step": 6731
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.369268951254475e-05,
+      "loss": 16.576,
+      "step": 6732
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.369050482930293e-05,
+      "loss": 14.8732,
+      "step": 6733
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.36883197932486e-05,
+      "loss": 14.9571,
+      "step": 6734
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.368613440439943e-05,
+      "loss": 17.9434,
+      "step": 6735
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.368394866277306e-05,
+      "loss": 15.9148,
+      "step": 6736
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.368176256838712e-05,
+      "loss": 16.5728,
+      "step": 6737
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.367957612125928e-05,
+      "loss": 16.8655,
+      "step": 6738
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.367738932140722e-05,
+      "loss": 14.8777,
+      "step": 6739
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.367520216884856e-05,
+      "loss": 15.3747,
+      "step": 6740
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.367301466360096e-05,
+      "loss": 17.4069,
+      "step": 6741
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.367082680568211e-05,
+      "loss": 14.2909,
+      "step": 6742
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.366863859510968e-05,
+      "loss": 15.165,
+      "step": 6743
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.366645003190132e-05,
+      "loss": 15.1676,
+      "step": 6744
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.366426111607472e-05,
+      "loss": 16.0869,
+      "step": 6745
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.366207184764753e-05,
+      "loss": 14.7359,
+      "step": 6746
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.365988222663747e-05,
+      "loss": 16.2392,
+      "step": 6747
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.365769225306218e-05,
+      "loss": 14.4409,
+      "step": 6748
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.365550192693937e-05,
+      "loss": 16.4483,
+      "step": 6749
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.365331124828671e-05,
+      "loss": 16.0891,
+      "step": 6750
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.365112021712192e-05,
+      "loss": 16.4814,
+      "step": 6751
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.364892883346265e-05,
+      "loss": 16.6338,
+      "step": 6752
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.364673709732665e-05,
+      "loss": 17.2083,
+      "step": 6753
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.364454500873157e-05,
+      "loss": 15.694,
+      "step": 6754
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.364235256769514e-05,
+      "loss": 14.405,
+      "step": 6755
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.364015977423504e-05,
+      "loss": 15.9667,
+      "step": 6756
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.363796662836901e-05,
+      "loss": 19.0729,
+      "step": 6757
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.363577313011473e-05,
+      "loss": 16.2476,
+      "step": 6758
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.363357927948994e-05,
+      "loss": 18.6761,
+      "step": 6759
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.363138507651233e-05,
+      "loss": 17.7645,
+      "step": 6760
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.362919052119963e-05,
+      "loss": 17.6709,
+      "step": 6761
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.362699561356956e-05,
+      "loss": 16.6725,
+      "step": 6762
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.362480035363986e-05,
+      "loss": 15.1696,
+      "step": 6763
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.362260474142823e-05,
+      "loss": 15.8486,
+      "step": 6764
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.362040877695243e-05,
+      "loss": 15.3209,
+      "step": 6765
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.361821246023017e-05,
+      "loss": 15.2773,
+      "step": 6766
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.361601579127919e-05,
+      "loss": 13.9283,
+      "step": 6767
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.361381877011723e-05,
+      "loss": 14.8877,
+      "step": 6768
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.361162139676204e-05,
+      "loss": 16.214,
+      "step": 6769
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.360942367123136e-05,
+      "loss": 17.1947,
+      "step": 6770
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.360722559354292e-05,
+      "loss": 16.1689,
+      "step": 6771
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.36050271637145e-05,
+      "loss": 15.7863,
+      "step": 6772
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.360282838176384e-05,
+      "loss": 15.7599,
+      "step": 6773
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.360062924770867e-05,
+      "loss": 16.3994,
+      "step": 6774
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.35984297615668e-05,
+      "loss": 15.6182,
+      "step": 6775
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.359622992335595e-05,
+      "loss": 16.8926,
+      "step": 6776
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.35940297330939e-05,
+      "loss": 16.9259,
+      "step": 6777
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.359182919079841e-05,
+      "loss": 16.613,
+      "step": 6778
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.358962829648727e-05,
+      "loss": 17.1146,
+      "step": 6779
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.358742705017822e-05,
+      "loss": 15.5753,
+      "step": 6780
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.358522545188906e-05,
+      "loss": 15.227,
+      "step": 6781
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.358302350163757e-05,
+      "loss": 16.5652,
+      "step": 6782
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.358082119944152e-05,
+      "loss": 13.8462,
+      "step": 6783
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.357861854531869e-05,
+      "loss": 17.6077,
+      "step": 6784
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.357641553928687e-05,
+      "loss": 15.0792,
+      "step": 6785
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.357421218136386e-05,
+      "loss": 15.5483,
+      "step": 6786
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.357200847156746e-05,
+      "loss": 15.1728,
+      "step": 6787
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.356980440991544e-05,
+      "loss": 17.5266,
+      "step": 6788
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.356759999642561e-05,
+      "loss": 16.3761,
+      "step": 6789
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.356539523111578e-05,
+      "loss": 17.2241,
+      "step": 6790
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.356319011400373e-05,
+      "loss": 14.921,
+      "step": 6791
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.35609846451073e-05,
+      "loss": 19.4857,
+      "step": 6792
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.355877882444428e-05,
+      "loss": 15.5384,
+      "step": 6793
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.355657265203249e-05,
+      "loss": 13.9769,
+      "step": 6794
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.355436612788973e-05,
+      "loss": 16.497,
+      "step": 6795
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.355215925203384e-05,
+      "loss": 17.7389,
+      "step": 6796
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.354995202448263e-05,
+      "loss": 16.7657,
+      "step": 6797
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.354774444525391e-05,
+      "loss": 15.5627,
+      "step": 6798
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.354553651436554e-05,
+      "loss": 15.2972,
+      "step": 6799
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.354332823183532e-05,
+      "loss": 17.1446,
+      "step": 6800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.354111959768109e-05,
+      "loss": 14.1949,
+      "step": 6801
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.353891061192069e-05,
+      "loss": 15.5291,
+      "step": 6802
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.353670127457198e-05,
+      "loss": 15.7704,
+      "step": 6803
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.353449158565273e-05,
+      "loss": 14.9779,
+      "step": 6804
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.353228154518087e-05,
+      "loss": 16.3205,
+      "step": 6805
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.353007115317421e-05,
+      "loss": 16.8427,
+      "step": 6806
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.352786040965058e-05,
+      "loss": 17.2715,
+      "step": 6807
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.352564931462785e-05,
+      "loss": 15.4159,
+      "step": 6808
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.352343786812387e-05,
+      "loss": 15.9889,
+      "step": 6809
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.352122607015652e-05,
+      "loss": 15.252,
+      "step": 6810
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.351901392074364e-05,
+      "loss": 17.8178,
+      "step": 6811
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.351680141990309e-05,
+      "loss": 15.6308,
+      "step": 6812
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.351458856765274e-05,
+      "loss": 15.5282,
+      "step": 6813
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.351237536401049e-05,
+      "loss": 15.5738,
+      "step": 6814
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.351016180899417e-05,
+      "loss": 17.0306,
+      "step": 6815
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.350794790262167e-05,
+      "loss": 14.2843,
+      "step": 6816
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.350573364491086e-05,
+      "loss": 14.6589,
+      "step": 6817
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.350351903587964e-05,
+      "loss": 16.1175,
+      "step": 6818
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.350130407554588e-05,
+      "loss": 16.3046,
+      "step": 6819
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.349908876392748e-05,
+      "loss": 15.9897,
+      "step": 6820
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.34968731010423e-05,
+      "loss": 16.8766,
+      "step": 6821
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.349465708690826e-05,
+      "loss": 15.8315,
+      "step": 6822
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.349244072154325e-05,
+      "loss": 15.8359,
+      "step": 6823
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.349022400496517e-05,
+      "loss": 16.0415,
+      "step": 6824
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.348800693719188e-05,
+      "loss": 19.4303,
+      "step": 6825
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.348578951824136e-05,
+      "loss": 15.0009,
+      "step": 6826
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.348357174813145e-05,
+      "loss": 16.3172,
+      "step": 6827
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.348135362688009e-05,
+      "loss": 15.4083,
+      "step": 6828
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.347913515450519e-05,
+      "loss": 15.9952,
+      "step": 6829
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.347691633102465e-05,
+      "loss": 19.3341,
+      "step": 6830
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.347469715645639e-05,
+      "loss": 15.5907,
+      "step": 6831
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.347247763081835e-05,
+      "loss": 17.1382,
+      "step": 6832
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.347025775412844e-05,
+      "loss": 14.9143,
+      "step": 6833
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.346803752640458e-05,
+      "loss": 15.3599,
+      "step": 6834
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.346581694766472e-05,
+      "loss": 16.028,
+      "step": 6835
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.346359601792675e-05,
+      "loss": 16.3473,
+      "step": 6836
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.346137473720866e-05,
+      "loss": 16.3914,
+      "step": 6837
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.345915310552835e-05,
+      "loss": 16.1155,
+      "step": 6838
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.345693112290376e-05,
+      "loss": 16.7488,
+      "step": 6839
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.345470878935286e-05,
+      "loss": 14.9754,
+      "step": 6840
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.345248610489356e-05,
+      "loss": 14.8607,
+      "step": 6841
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.345026306954386e-05,
+      "loss": 16.6054,
+      "step": 6842
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.344803968332166e-05,
+      "loss": 16.4925,
+      "step": 6843
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.344581594624492e-05,
+      "loss": 15.1497,
+      "step": 6844
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.344359185833162e-05,
+      "loss": 18.7865,
+      "step": 6845
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.344136741959971e-05,
+      "loss": 15.668,
+      "step": 6846
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.343914263006716e-05,
+      "loss": 18.0175,
+      "step": 6847
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.343691748975193e-05,
+      "loss": 16.8022,
+      "step": 6848
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.343469199867199e-05,
+      "loss": 17.1101,
+      "step": 6849
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.34324661568453e-05,
+      "loss": 16.8088,
+      "step": 6850
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.343023996428986e-05,
+      "loss": 16.7819,
+      "step": 6851
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.342801342102361e-05,
+      "loss": 16.8492,
+      "step": 6852
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.342578652706457e-05,
+      "loss": 15.5378,
+      "step": 6853
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.34235592824307e-05,
+      "loss": 15.4455,
+      "step": 6854
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.342133168713998e-05,
+      "loss": 17.4754,
+      "step": 6855
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.341910374121041e-05,
+      "loss": 17.9537,
+      "step": 6856
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.341687544465998e-05,
+      "loss": 15.9021,
+      "step": 6857
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.341464679750669e-05,
+      "loss": 16.9819,
+      "step": 6858
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.341241779976853e-05,
+      "loss": 18.4775,
+      "step": 6859
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.34101884514635e-05,
+      "loss": 15.9589,
+      "step": 6860
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.340795875260959e-05,
+      "loss": 15.4067,
+      "step": 6861
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.340572870322483e-05,
+      "loss": 15.5535,
+      "step": 6862
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.34034983033272e-05,
+      "loss": 16.1672,
+      "step": 6863
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.340126755293475e-05,
+      "loss": 17.2662,
+      "step": 6864
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.339903645206546e-05,
+      "loss": 15.2891,
+      "step": 6865
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.339680500073736e-05,
+      "loss": 16.6365,
+      "step": 6866
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.339457319896847e-05,
+      "loss": 16.5394,
+      "step": 6867
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.339234104677681e-05,
+      "loss": 16.6709,
+      "step": 6868
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.33901085441804e-05,
+      "loss": 14.0622,
+      "step": 6869
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.338787569119727e-05,
+      "loss": 17.3479,
+      "step": 6870
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.338564248784547e-05,
+      "loss": 16.4474,
+      "step": 6871
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.338340893414299e-05,
+      "loss": 16.0982,
+      "step": 6872
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.338117503010793e-05,
+      "loss": 15.3429,
+      "step": 6873
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.337894077575827e-05,
+      "loss": 14.9752,
+      "step": 6874
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.337670617111207e-05,
+      "loss": 15.3027,
+      "step": 6875
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.337447121618738e-05,
+      "loss": 16.0511,
+      "step": 6876
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.337223591100228e-05,
+      "loss": 17.3049,
+      "step": 6877
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.337000025557476e-05,
+      "loss": 17.4481,
+      "step": 6878
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.336776424992291e-05,
+      "loss": 17.6035,
+      "step": 6879
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.336552789406479e-05,
+      "loss": 19.3117,
+      "step": 6880
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.336329118801843e-05,
+      "loss": 15.8229,
+      "step": 6881
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.336105413180192e-05,
+      "loss": 17.4868,
+      "step": 6882
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.335881672543332e-05,
+      "loss": 16.5978,
+      "step": 6883
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.335657896893069e-05,
+      "loss": 15.5326,
+      "step": 6884
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.33543408623121e-05,
+      "loss": 17.0961,
+      "step": 6885
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.335210240559564e-05,
+      "loss": 14.9545,
+      "step": 6886
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.334986359879936e-05,
+      "loss": 16.0559,
+      "step": 6887
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.334762444194135e-05,
+      "loss": 16.8423,
+      "step": 6888
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.33453849350397e-05,
+      "loss": 16.2098,
+      "step": 6889
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.334314507811248e-05,
+      "loss": 15.8622,
+      "step": 6890
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.33409048711778e-05,
+      "loss": 17.4289,
+      "step": 6891
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.333866431425372e-05,
+      "loss": 16.0183,
+      "step": 6892
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.333642340735836e-05,
+      "loss": 18.0151,
+      "step": 6893
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.33341821505098e-05,
+      "loss": 16.458,
+      "step": 6894
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.333194054372613e-05,
+      "loss": 14.9913,
+      "step": 6895
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.33296985870255e-05,
+      "loss": 16.2504,
+      "step": 6896
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.332745628042595e-05,
+      "loss": 17.4796,
+      "step": 6897
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.332521362394563e-05,
+      "loss": 16.1907,
+      "step": 6898
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.332297061760263e-05,
+      "loss": 15.8683,
+      "step": 6899
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.332072726141508e-05,
+      "loss": 15.1292,
+      "step": 6900
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.331848355540107e-05,
+      "loss": 15.0281,
+      "step": 6901
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.331623949957875e-05,
+      "loss": 16.4255,
+      "step": 6902
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.331399509396622e-05,
+      "loss": 15.9936,
+      "step": 6903
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.33117503385816e-05,
+      "loss": 16.0519,
+      "step": 6904
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.330950523344304e-05,
+      "loss": 19.5444,
+      "step": 6905
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.330725977856864e-05,
+      "loss": 15.8269,
+      "step": 6906
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.330501397397657e-05,
+      "loss": 15.769,
+      "step": 6907
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.330276781968493e-05,
+      "loss": 17.8896,
+      "step": 6908
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.330052131571188e-05,
+      "loss": 15.2641,
+      "step": 6909
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.329827446207556e-05,
+      "loss": 16.221,
+      "step": 6910
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.32960272587941e-05,
+      "loss": 17.7172,
+      "step": 6911
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.329377970588565e-05,
+      "loss": 14.9214,
+      "step": 6912
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.329153180336836e-05,
+      "loss": 16.312,
+      "step": 6913
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.32892835512604e-05,
+      "loss": 16.3823,
+      "step": 6914
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.328703494957991e-05,
+      "loss": 16.0675,
+      "step": 6915
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.328478599834505e-05,
+      "loss": 17.0665,
+      "step": 6916
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.328253669757396e-05,
+      "loss": 16.2011,
+      "step": 6917
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.328028704728486e-05,
+      "loss": 16.73,
+      "step": 6918
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.327803704749585e-05,
+      "loss": 15.2298,
+      "step": 6919
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.327578669822515e-05,
+      "loss": 15.2041,
+      "step": 6920
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.327353599949091e-05,
+      "loss": 17.2503,
+      "step": 6921
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.32712849513113e-05,
+      "loss": 15.5489,
+      "step": 6922
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.326903355370451e-05,
+      "loss": 16.3671,
+      "step": 6923
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.326678180668871e-05,
+      "loss": 15.1866,
+      "step": 6924
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.32645297102821e-05,
+      "loss": 15.6359,
+      "step": 6925
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.326227726450285e-05,
+      "loss": 14.7875,
+      "step": 6926
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.326002446936914e-05,
+      "loss": 15.9868,
+      "step": 6927
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.32577713248992e-05,
+      "loss": 17.2963,
+      "step": 6928
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.325551783111119e-05,
+      "loss": 14.8643,
+      "step": 6929
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.325326398802333e-05,
+      "loss": 17.8733,
+      "step": 6930
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.32510097956538e-05,
+      "loss": 17.5327,
+      "step": 6931
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.324875525402081e-05,
+      "loss": 16.3074,
+      "step": 6932
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.324650036314257e-05,
+      "loss": 16.1548,
+      "step": 6933
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.32442451230373e-05,
+      "loss": 15.277,
+      "step": 6934
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.324198953372319e-05,
+      "loss": 16.67,
+      "step": 6935
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.323973359521847e-05,
+      "loss": 17.7341,
+      "step": 6936
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.323747730754134e-05,
+      "loss": 15.5229,
+      "step": 6937
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.323522067071004e-05,
+      "loss": 14.9677,
+      "step": 6938
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.323296368474278e-05,
+      "loss": 16.6241,
+      "step": 6939
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.323070634965779e-05,
+      "loss": 16.3849,
+      "step": 6940
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.32284486654733e-05,
+      "loss": 16.0424,
+      "step": 6941
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.322619063220755e-05,
+      "loss": 17.3037,
+      "step": 6942
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.322393224987875e-05,
+      "loss": 17.5214,
+      "step": 6943
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.322167351850515e-05,
+      "loss": 18.1028,
+      "step": 6944
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.321941443810499e-05,
+      "loss": 18.5643,
+      "step": 6945
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.321715500869652e-05,
+      "loss": 17.3947,
+      "step": 6946
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.321489523029797e-05,
+      "loss": 16.0752,
+      "step": 6947
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.321263510292762e-05,
+      "loss": 17.7808,
+      "step": 6948
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.321037462660367e-05,
+      "loss": 16.3039,
+      "step": 6949
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.320811380134443e-05,
+      "loss": 17.0964,
+      "step": 6950
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.32058526271681e-05,
+      "loss": 16.3198,
+      "step": 6951
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.320359110409298e-05,
+      "loss": 14.2445,
+      "step": 6952
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.320132923213734e-05,
+      "loss": 16.8194,
+      "step": 6953
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.31990670113194e-05,
+      "loss": 15.7397,
+      "step": 6954
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.319680444165747e-05,
+      "loss": 17.9985,
+      "step": 6955
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.31945415231698e-05,
+      "loss": 16.0478,
+      "step": 6956
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.319227825587467e-05,
+      "loss": 16.9928,
+      "step": 6957
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.319001463979036e-05,
+      "loss": 18.6119,
+      "step": 6958
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.318775067493514e-05,
+      "loss": 16.5475,
+      "step": 6959
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.31854863613273e-05,
+      "loss": 16.613,
+      "step": 6960
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.318322169898513e-05,
+      "loss": 15.894,
+      "step": 6961
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.318095668792689e-05,
+      "loss": 17.1605,
+      "step": 6962
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.317869132817089e-05,
+      "loss": 17.8465,
+      "step": 6963
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.317642561973545e-05,
+      "loss": 16.2175,
+      "step": 6964
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.317415956263882e-05,
+      "loss": 18.2662,
+      "step": 6965
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.317189315689932e-05,
+      "loss": 18.0757,
+      "step": 6966
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.316962640253526e-05,
+      "loss": 16.0238,
+      "step": 6967
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.316735929956492e-05,
+      "loss": 15.7211,
+      "step": 6968
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.316509184800664e-05,
+      "loss": 16.6084,
+      "step": 6969
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.316282404787871e-05,
+      "loss": 16.1093,
+      "step": 6970
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.316055589919943e-05,
+      "loss": 14.6408,
+      "step": 6971
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.315828740198714e-05,
+      "loss": 16.3059,
+      "step": 6972
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.315601855626015e-05,
+      "loss": 15.2254,
+      "step": 6973
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.315374936203679e-05,
+      "loss": 17.0074,
+      "step": 6974
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.315147981933536e-05,
+      "loss": 16.0746,
+      "step": 6975
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.314920992817424e-05,
+      "loss": 14.7826,
+      "step": 6976
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.314693968857168e-05,
+      "loss": 16.5883,
+      "step": 6977
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.314466910054609e-05,
+      "loss": 15.1397,
+      "step": 6978
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.314239816411576e-05,
+      "loss": 15.2721,
+      "step": 6979
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.314012687929902e-05,
+      "loss": 15.6363,
+      "step": 6980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.313785524611427e-05,
+      "loss": 14.8343,
+      "step": 6981
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.313558326457979e-05,
+      "loss": 16.9624,
+      "step": 6982
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.313331093471396e-05,
+      "loss": 16.2446,
+      "step": 6983
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.313103825653513e-05,
+      "loss": 16.1291,
+      "step": 6984
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.312876523006163e-05,
+      "loss": 15.2679,
+      "step": 6985
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.312649185531184e-05,
+      "loss": 15.8864,
+      "step": 6986
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.312421813230411e-05,
+      "loss": 15.6197,
+      "step": 6987
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.312194406105679e-05,
+      "loss": 16.4198,
+      "step": 6988
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.311966964158827e-05,
+      "loss": 17.4444,
+      "step": 6989
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.311739487391689e-05,
+      "loss": 17.5025,
+      "step": 6990
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.311511975806102e-05,
+      "loss": 15.3351,
+      "step": 6991
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.311284429403903e-05,
+      "loss": 15.973,
+      "step": 6992
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.311056848186935e-05,
+      "loss": 16.5365,
+      "step": 6993
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.310829232157027e-05,
+      "loss": 15.9469,
+      "step": 6994
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.310601581316023e-05,
+      "loss": 17.0555,
+      "step": 6995
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.31037389566576e-05,
+      "loss": 15.2578,
+      "step": 6996
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.310146175208076e-05,
+      "loss": 17.3392,
+      "step": 6997
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.309918419944812e-05,
+      "loss": 16.3417,
+      "step": 6998
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.309690629877803e-05,
+      "loss": 15.4904,
+      "step": 6999
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.309462805008893e-05,
+      "loss": 15.8744,
+      "step": 7000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.309234945339918e-05,
+      "loss": 18.3417,
+      "step": 7001
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.309007050872721e-05,
+      "loss": 15.2863,
+      "step": 7002
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.308779121609141e-05,
+      "loss": 15.6905,
+      "step": 7003
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.308551157551018e-05,
+      "loss": 14.9174,
+      "step": 7004
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.308323158700194e-05,
+      "loss": 16.9045,
+      "step": 7005
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.30809512505851e-05,
+      "loss": 16.0001,
+      "step": 7006
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.307867056627806e-05,
+      "loss": 17.0846,
+      "step": 7007
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.307638953409925e-05,
+      "loss": 15.1161,
+      "step": 7008
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.307410815406709e-05,
+      "loss": 13.7898,
+      "step": 7009
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.30718264262e-05,
+      "loss": 17.3052,
+      "step": 7010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.306954435051642e-05,
+      "loss": 14.0679,
+      "step": 7011
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.306726192703474e-05,
+      "loss": 17.0284,
+      "step": 7012
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.306497915577343e-05,
+      "loss": 16.3554,
+      "step": 7013
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.306269603675091e-05,
+      "loss": 15.3085,
+      "step": 7014
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.306041256998561e-05,
+      "loss": 17.1368,
+      "step": 7015
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.305812875549599e-05,
+      "loss": 16.0077,
+      "step": 7016
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.305584459330047e-05,
+      "loss": 17.6626,
+      "step": 7017
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.30535600834175e-05,
+      "loss": 14.3664,
+      "step": 7018
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.305127522586554e-05,
+      "loss": 16.2667,
+      "step": 7019
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.304899002066303e-05,
+      "loss": 15.4175,
+      "step": 7020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.304670446782842e-05,
+      "loss": 16.7946,
+      "step": 7021
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.304441856738018e-05,
+      "loss": 17.8525,
+      "step": 7022
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.304213231933676e-05,
+      "loss": 16.8427,
+      "step": 7023
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.303984572371662e-05,
+      "loss": 15.5422,
+      "step": 7024
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.303755878053822e-05,
+      "loss": 17.0909,
+      "step": 7025
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.303527148982005e-05,
+      "loss": 16.3398,
+      "step": 7026
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.303298385158057e-05,
+      "loss": 16.0767,
+      "step": 7027
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.303069586583825e-05,
+      "loss": 18.3913,
+      "step": 7028
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.302840753261157e-05,
+      "loss": 16.5748,
+      "step": 7029
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.302611885191899e-05,
+      "loss": 15.787,
+      "step": 7030
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.302382982377902e-05,
+      "loss": 15.7643,
+      "step": 7031
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.302154044821013e-05,
+      "loss": 15.5431,
+      "step": 7032
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.30192507252308e-05,
+      "loss": 15.7267,
+      "step": 7033
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.301696065485954e-05,
+      "loss": 15.5478,
+      "step": 7034
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.301467023711483e-05,
+      "loss": 16.1521,
+      "step": 7035
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.301237947201515e-05,
+      "loss": 16.0141,
+      "step": 7036
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.301008835957905e-05,
+      "loss": 17.9839,
+      "step": 7037
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.300779689982498e-05,
+      "loss": 15.6445,
+      "step": 7038
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.300550509277145e-05,
+      "loss": 15.0969,
+      "step": 7039
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.300321293843699e-05,
+      "loss": 16.9688,
+      "step": 7040
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.30009204368401e-05,
+      "loss": 15.6292,
+      "step": 7041
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.299862758799929e-05,
+      "loss": 16.9925,
+      "step": 7042
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.299633439193308e-05,
+      "loss": 17.7085,
+      "step": 7043
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.299404084865997e-05,
+      "loss": 16.8037,
+      "step": 7044
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.29917469581985e-05,
+      "loss": 16.7741,
+      "step": 7045
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.298945272056721e-05,
+      "loss": 16.7188,
+      "step": 7046
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.298715813578457e-05,
+      "loss": 18.3362,
+      "step": 7047
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.298486320386919e-05,
+      "loss": 17.758,
+      "step": 7048
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.298256792483952e-05,
+      "loss": 16.8817,
+      "step": 7049
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.298027229871416e-05,
+      "loss": 16.6846,
+      "step": 7050
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.29779763255116e-05,
+      "loss": 16.2373,
+      "step": 7051
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.297568000525042e-05,
+      "loss": 16.536,
+      "step": 7052
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.297338333794912e-05,
+      "loss": 15.7458,
+      "step": 7053
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.297108632362628e-05,
+      "loss": 17.758,
+      "step": 7054
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.296878896230043e-05,
+      "loss": 17.3893,
+      "step": 7055
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.296649125399015e-05,
+      "loss": 14.5801,
+      "step": 7056
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.296419319871397e-05,
+      "loss": 16.3211,
+      "step": 7057
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.296189479649046e-05,
+      "loss": 17.5212,
+      "step": 7058
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.295959604733816e-05,
+      "loss": 16.119,
+      "step": 7059
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.295729695127565e-05,
+      "loss": 17.5137,
+      "step": 7060
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.295499750832151e-05,
+      "loss": 16.3875,
+      "step": 7061
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.295269771849427e-05,
+      "loss": 13.8608,
+      "step": 7062
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.295039758181252e-05,
+      "loss": 17.3918,
+      "step": 7063
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.294809709829484e-05,
+      "loss": 16.84,
+      "step": 7064
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.294579626795981e-05,
+      "loss": 18.327,
+      "step": 7065
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.294349509082601e-05,
+      "loss": 15.5245,
+      "step": 7066
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.2941193566912e-05,
+      "loss": 15.825,
+      "step": 7067
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.293889169623639e-05,
+      "loss": 15.8887,
+      "step": 7068
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.293658947881776e-05,
+      "loss": 16.1644,
+      "step": 7069
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.29342869146747e-05,
+      "loss": 16.5825,
+      "step": 7070
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.293198400382582e-05,
+      "loss": 17.0714,
+      "step": 7071
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.292968074628967e-05,
+      "loss": 17.4065,
+      "step": 7072
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.292737714208491e-05,
+      "loss": 18.1754,
+      "step": 7073
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.292507319123009e-05,
+      "loss": 15.4961,
+      "step": 7074
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.292276889374386e-05,
+      "loss": 16.3187,
+      "step": 7075
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.292046424964477e-05,
+      "loss": 17.5662,
+      "step": 7076
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.291815925895149e-05,
+      "loss": 15.163,
+      "step": 7077
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.291585392168262e-05,
+      "loss": 17.1719,
+      "step": 7078
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.291354823785675e-05,
+      "loss": 15.4088,
+      "step": 7079
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.291124220749253e-05,
+      "loss": 16.1012,
+      "step": 7080
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.290893583060855e-05,
+      "loss": 16.8996,
+      "step": 7081
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.290662910722345e-05,
+      "loss": 15.8272,
+      "step": 7082
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.290432203735588e-05,
+      "loss": 15.4684,
+      "step": 7083
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.290201462102443e-05,
+      "loss": 17.0733,
+      "step": 7084
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.289970685824776e-05,
+      "loss": 17.6205,
+      "step": 7085
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.289739874904449e-05,
+      "loss": 16.0784,
+      "step": 7086
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.289509029343326e-05,
+      "loss": 15.0356,
+      "step": 7087
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.289278149143273e-05,
+      "loss": 18.1163,
+      "step": 7088
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.289047234306154e-05,
+      "loss": 16.0607,
+      "step": 7089
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.288816284833833e-05,
+      "loss": 16.109,
+      "step": 7090
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.288585300728171e-05,
+      "loss": 15.9958,
+      "step": 7091
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.288354281991042e-05,
+      "loss": 16.9045,
+      "step": 7092
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.288123228624304e-05,
+      "loss": 16.0632,
+      "step": 7093
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.287892140629825e-05,
+      "loss": 17.0162,
+      "step": 7094
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.287661018009475e-05,
+      "loss": 15.7099,
+      "step": 7095
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.287429860765113e-05,
+      "loss": 17.8937,
+      "step": 7096
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.287198668898613e-05,
+      "loss": 16.5255,
+      "step": 7097
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.286967442411835e-05,
+      "loss": 16.6085,
+      "step": 7098
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.286736181306652e-05,
+      "loss": 15.1793,
+      "step": 7099
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.28650488558493e-05,
+      "loss": 16.2163,
+      "step": 7100
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.286273555248534e-05,
+      "loss": 17.934,
+      "step": 7101
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.286042190299335e-05,
+      "loss": 17.9043,
+      "step": 7102
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.285810790739202e-05,
+      "loss": 15.082,
+      "step": 7103
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.28557935657e-05,
+      "loss": 18.0266,
+      "step": 7104
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.285347887793599e-05,
+      "loss": 16.5306,
+      "step": 7105
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.285116384411872e-05,
+      "loss": 15.8562,
+      "step": 7106
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.284884846426684e-05,
+      "loss": 15.1047,
+      "step": 7107
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.284653273839905e-05,
+      "loss": 16.1391,
+      "step": 7108
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.284421666653409e-05,
+      "loss": 16.1125,
+      "step": 7109
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.284190024869062e-05,
+      "loss": 17.7004,
+      "step": 7110
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.283958348488734e-05,
+      "loss": 15.7184,
+      "step": 7111
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.283726637514303e-05,
+      "loss": 17.4756,
+      "step": 7112
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.283494891947632e-05,
+      "loss": 17.0085,
+      "step": 7113
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.283263111790596e-05,
+      "loss": 16.6662,
+      "step": 7114
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.283031297045067e-05,
+      "loss": 17.0825,
+      "step": 7115
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.282799447712916e-05,
+      "loss": 16.8246,
+      "step": 7116
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.282567563796014e-05,
+      "loss": 15.2818,
+      "step": 7117
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.282335645296236e-05,
+      "loss": 15.6612,
+      "step": 7118
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.282103692215454e-05,
+      "loss": 15.1185,
+      "step": 7119
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.281871704555541e-05,
+      "loss": 15.5736,
+      "step": 7120
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.281639682318369e-05,
+      "loss": 15.1144,
+      "step": 7121
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.281407625505813e-05,
+      "loss": 16.9619,
+      "step": 7122
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.281175534119748e-05,
+      "loss": 15.0691,
+      "step": 7123
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.280943408162046e-05,
+      "loss": 15.7497,
+      "step": 7124
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.280711247634583e-05,
+      "loss": 14.848,
+      "step": 7125
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.280479052539233e-05,
+      "loss": 16.9237,
+      "step": 7126
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.28024682287787e-05,
+      "loss": 16.1763,
+      "step": 7127
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.280014558652372e-05,
+      "loss": 16.6481,
+      "step": 7128
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.279782259864613e-05,
+      "loss": 15.6131,
+      "step": 7129
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.279549926516469e-05,
+      "loss": 15.1936,
+      "step": 7130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.279317558609817e-05,
+      "loss": 16.3053,
+      "step": 7131
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.279085156146532e-05,
+      "loss": 18.34,
+      "step": 7132
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.27885271912849e-05,
+      "loss": 16.801,
+      "step": 7133
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.278620247557571e-05,
+      "loss": 18.1462,
+      "step": 7134
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.27838774143565e-05,
+      "loss": 16.5389,
+      "step": 7135
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.278155200764606e-05,
+      "loss": 15.3847,
+      "step": 7136
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.277922625546315e-05,
+      "loss": 16.7668,
+      "step": 7137
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.277690015782657e-05,
+      "loss": 16.8414,
+      "step": 7138
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.277457371475508e-05,
+      "loss": 14.8002,
+      "step": 7139
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.277224692626749e-05,
+      "loss": 15.5795,
+      "step": 7140
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.276991979238258e-05,
+      "loss": 15.4205,
+      "step": 7141
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.276759231311914e-05,
+      "loss": 17.4601,
+      "step": 7142
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.276526448849596e-05,
+      "loss": 15.0693,
+      "step": 7143
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.276293631853185e-05,
+      "loss": 17.9608,
+      "step": 7144
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.276060780324563e-05,
+      "loss": 16.7146,
+      "step": 7145
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.275827894265605e-05,
+      "loss": 14.7787,
+      "step": 7146
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.275594973678196e-05,
+      "loss": 16.3988,
+      "step": 7147
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.275362018564215e-05,
+      "loss": 15.2564,
+      "step": 7148
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.275129028925545e-05,
+      "loss": 14.3579,
+      "step": 7149
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.274896004764064e-05,
+      "loss": 16.802,
+      "step": 7150
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.274662946081658e-05,
+      "loss": 15.6318,
+      "step": 7151
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.274429852880207e-05,
+      "loss": 15.5906,
+      "step": 7152
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.27419672516159e-05,
+      "loss": 17.6393,
+      "step": 7153
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.273963562927695e-05,
+      "loss": 16.885,
+      "step": 7154
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.273730366180403e-05,
+      "loss": 18.1583,
+      "step": 7155
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.273497134921596e-05,
+      "loss": 15.7472,
+      "step": 7156
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.273263869153157e-05,
+      "loss": 14.3209,
+      "step": 7157
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.273030568876972e-05,
+      "loss": 16.1735,
+      "step": 7158
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.272797234094925e-05,
+      "loss": 16.0093,
+      "step": 7159
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.272563864808897e-05,
+      "loss": 16.9207,
+      "step": 7160
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.272330461020774e-05,
+      "loss": 14.9374,
+      "step": 7161
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.272097022732443e-05,
+      "loss": 16.9386,
+      "step": 7162
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.271863549945787e-05,
+      "loss": 16.5695,
+      "step": 7163
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.27163004266269e-05,
+      "loss": 16.0094,
+      "step": 7164
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.271396500885042e-05,
+      "loss": 16.5191,
+      "step": 7165
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.271162924614725e-05,
+      "loss": 16.0925,
+      "step": 7166
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.270929313853627e-05,
+      "loss": 14.6365,
+      "step": 7167
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.270695668603634e-05,
+      "loss": 15.2775,
+      "step": 7168
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.270461988866632e-05,
+      "loss": 15.7594,
+      "step": 7169
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.27022827464451e-05,
+      "loss": 14.9754,
+      "step": 7170
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.269994525939151e-05,
+      "loss": 17.7305,
+      "step": 7171
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.26976074275245e-05,
+      "loss": 14.9308,
+      "step": 7172
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.269526925086288e-05,
+      "loss": 16.631,
+      "step": 7173
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.269293072942554e-05,
+      "loss": 17.3484,
+      "step": 7174
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.26905918632314e-05,
+      "loss": 14.2041,
+      "step": 7175
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.268825265229933e-05,
+      "loss": 16.9422,
+      "step": 7176
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.26859130966482e-05,
+      "loss": 15.1665,
+      "step": 7177
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.268357319629693e-05,
+      "loss": 15.3148,
+      "step": 7178
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.26812329512644e-05,
+      "loss": 14.7296,
+      "step": 7179
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.267889236156952e-05,
+      "loss": 17.2366,
+      "step": 7180
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.267655142723119e-05,
+      "loss": 17.3119,
+      "step": 7181
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.26742101482683e-05,
+      "loss": 14.6356,
+      "step": 7182
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.267186852469976e-05,
+      "loss": 14.434,
+      "step": 7183
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.266952655654448e-05,
+      "loss": 16.5509,
+      "step": 7184
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.26671842438214e-05,
+      "loss": 16.4206,
+      "step": 7185
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.266484158654938e-05,
+      "loss": 14.9006,
+      "step": 7186
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.266249858474738e-05,
+      "loss": 17.2399,
+      "step": 7187
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.26601552384343e-05,
+      "loss": 15.8905,
+      "step": 7188
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.265781154762908e-05,
+      "loss": 14.8814,
+      "step": 7189
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.265546751235064e-05,
+      "loss": 16.1996,
+      "step": 7190
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.26531231326179e-05,
+      "loss": 15.5793,
+      "step": 7191
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.26507784084498e-05,
+      "loss": 16.1232,
+      "step": 7192
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.264843333986526e-05,
+      "loss": 16.59,
+      "step": 7193
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.264608792688326e-05,
+      "loss": 16.6572,
+      "step": 7194
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.264374216952267e-05,
+      "loss": 16.3079,
+      "step": 7195
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.264139606780249e-05,
+      "loss": 15.7658,
+      "step": 7196
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.263904962174163e-05,
+      "loss": 16.0947,
+      "step": 7197
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.263670283135908e-05,
+      "loss": 16.7841,
+      "step": 7198
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.263435569667376e-05,
+      "loss": 18.8872,
+      "step": 7199
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.263200821770461e-05,
+      "loss": 15.7706,
+      "step": 7200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.262966039447063e-05,
+      "loss": 14.7072,
+      "step": 7201
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.262731222699074e-05,
+      "loss": 16.1598,
+      "step": 7202
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.262496371528393e-05,
+      "loss": 16.2891,
+      "step": 7203
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.262261485936914e-05,
+      "loss": 16.4933,
+      "step": 7204
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.262026565926534e-05,
+      "loss": 18.1905,
+      "step": 7205
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.261791611499154e-05,
+      "loss": 17.9958,
+      "step": 7206
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.261556622656666e-05,
+      "loss": 17.0094,
+      "step": 7207
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.26132159940097e-05,
+      "loss": 16.3443,
+      "step": 7208
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.261086541733965e-05,
+      "loss": 15.4892,
+      "step": 7209
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.260851449657546e-05,
+      "loss": 16.0795,
+      "step": 7210
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.260616323173615e-05,
+      "loss": 17.5324,
+      "step": 7211
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.260381162284068e-05,
+      "loss": 14.6806,
+      "step": 7212
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.260145966990807e-05,
+      "loss": 14.2163,
+      "step": 7213
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.259910737295729e-05,
+      "loss": 17.2052,
+      "step": 7214
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.259675473200732e-05,
+      "loss": 16.0239,
+      "step": 7215
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.259440174707719e-05,
+      "loss": 16.6397,
+      "step": 7216
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.259204841818588e-05,
+      "loss": 15.5808,
+      "step": 7217
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.258969474535241e-05,
+      "loss": 16.6514,
+      "step": 7218
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.258734072859578e-05,
+      "loss": 14.6952,
+      "step": 7219
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.2584986367935e-05,
+      "loss": 16.7558,
+      "step": 7220
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.258263166338908e-05,
+      "loss": 15.5412,
+      "step": 7221
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.258027661497702e-05,
+      "loss": 14.2917,
+      "step": 7222
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.257792122271785e-05,
+      "loss": 14.4039,
+      "step": 7223
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.257556548663061e-05,
+      "loss": 16.7851,
+      "step": 7224
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.257320940673431e-05,
+      "loss": 14.6974,
+      "step": 7225
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.257085298304795e-05,
+      "loss": 15.648,
+      "step": 7226
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.256849621559059e-05,
+      "loss": 16.179,
+      "step": 7227
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.256613910438126e-05,
+      "loss": 16.1294,
+      "step": 7228
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.256378164943897e-05,
+      "loss": 15.1774,
+      "step": 7229
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.25614238507828e-05,
+      "loss": 16.3399,
+      "step": 7230
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.255906570843173e-05,
+      "loss": 15.0261,
+      "step": 7231
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.255670722240487e-05,
+      "loss": 15.3166,
+      "step": 7232
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.255434839272122e-05,
+      "loss": 16.3546,
+      "step": 7233
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.255198921939982e-05,
+      "loss": 17.37,
+      "step": 7234
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.254962970245976e-05,
+      "loss": 14.3739,
+      "step": 7235
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.254726984192006e-05,
+      "loss": 16.5997,
+      "step": 7236
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.25449096377998e-05,
+      "loss": 16.0861,
+      "step": 7237
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.254254909011804e-05,
+      "loss": 16.138,
+      "step": 7238
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.254018819889381e-05,
+      "loss": 14.74,
+      "step": 7239
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.253782696414621e-05,
+      "loss": 15.6527,
+      "step": 7240
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.25354653858943e-05,
+      "loss": 15.8908,
+      "step": 7241
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.253310346415714e-05,
+      "loss": 14.4855,
+      "step": 7242
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.253074119895381e-05,
+      "loss": 15.552,
+      "step": 7243
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.252837859030337e-05,
+      "loss": 15.8303,
+      "step": 7244
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.252601563822493e-05,
+      "loss": 15.7822,
+      "step": 7245
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.252365234273755e-05,
+      "loss": 14.8512,
+      "step": 7246
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.252128870386032e-05,
+      "loss": 15.2517,
+      "step": 7247
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.251892472161232e-05,
+      "loss": 15.7006,
+      "step": 7248
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.251656039601264e-05,
+      "loss": 15.5904,
+      "step": 7249
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.251419572708039e-05,
+      "loss": 16.6367,
+      "step": 7250
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.251183071483464e-05,
+      "loss": 16.5912,
+      "step": 7251
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.250946535929451e-05,
+      "loss": 17.928,
+      "step": 7252
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.25070996604791e-05,
+      "loss": 18.1275,
+      "step": 7253
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.250473361840748e-05,
+      "loss": 16.7873,
+      "step": 7254
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.250236723309882e-05,
+      "loss": 15.9708,
+      "step": 7255
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.250000050457216e-05,
+      "loss": 16.5656,
+      "step": 7256
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.249763343284665e-05,
+      "loss": 16.098,
+      "step": 7257
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.24952660179414e-05,
+      "loss": 16.4632,
+      "step": 7258
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.249289825987552e-05,
+      "loss": 16.5881,
+      "step": 7259
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.249053015866815e-05,
+      "loss": 17.1465,
+      "step": 7260
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.248816171433841e-05,
+      "loss": 16.7856,
+      "step": 7261
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.248579292690538e-05,
+      "loss": 16.0792,
+      "step": 7262
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.248342379638824e-05,
+      "loss": 16.5284,
+      "step": 7263
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.248105432280611e-05,
+      "loss": 17.8464,
+      "step": 7264
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.247868450617812e-05,
+      "loss": 17.8211,
+      "step": 7265
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.24763143465234e-05,
+      "loss": 17.2447,
+      "step": 7266
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.24739438438611e-05,
+      "loss": 16.7568,
+      "step": 7267
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.247157299821034e-05,
+      "loss": 15.4617,
+      "step": 7268
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.246920180959029e-05,
+      "loss": 15.5561,
+      "step": 7269
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.24668302780201e-05,
+      "loss": 17.257,
+      "step": 7270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.246445840351892e-05,
+      "loss": 14.944,
+      "step": 7271
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.246208618610588e-05,
+      "loss": 14.9184,
+      "step": 7272
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.245971362580015e-05,
+      "loss": 16.3189,
+      "step": 7273
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.245734072262089e-05,
+      "loss": 16.6895,
+      "step": 7274
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.245496747658727e-05,
+      "loss": 18.5317,
+      "step": 7275
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.245259388771845e-05,
+      "loss": 15.3412,
+      "step": 7276
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.24502199560336e-05,
+      "loss": 16.7097,
+      "step": 7277
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.244784568155186e-05,
+      "loss": 16.9655,
+      "step": 7278
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.244547106429245e-05,
+      "loss": 16.1727,
+      "step": 7279
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.24430961042745e-05,
+      "loss": 15.6938,
+      "step": 7280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.244072080151723e-05,
+      "loss": 14.4625,
+      "step": 7281
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.24383451560398e-05,
+      "loss": 15.3691,
+      "step": 7282
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.243596916786141e-05,
+      "loss": 16.9028,
+      "step": 7283
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.243359283700121e-05,
+      "loss": 15.4218,
+      "step": 7284
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.243121616347843e-05,
+      "loss": 16.1592,
+      "step": 7285
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.242883914731223e-05,
+      "loss": 17.0969,
+      "step": 7286
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.242646178852184e-05,
+      "loss": 15.4211,
+      "step": 7287
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.242408408712642e-05,
+      "loss": 14.8088,
+      "step": 7288
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.24217060431452e-05,
+      "loss": 17.4926,
+      "step": 7289
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.241932765659735e-05,
+      "loss": 16.0118,
+      "step": 7290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.241694892750212e-05,
+      "loss": 17.0992,
+      "step": 7291
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.241456985587868e-05,
+      "loss": 15.1146,
+      "step": 7292
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.241219044174628e-05,
+      "loss": 17.3568,
+      "step": 7293
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.240981068512411e-05,
+      "loss": 15.1513,
+      "step": 7294
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.240743058603139e-05,
+      "loss": 16.7266,
+      "step": 7295
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.240505014448732e-05,
+      "loss": 16.8682,
+      "step": 7296
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.240266936051117e-05,
+      "loss": 17.7787,
+      "step": 7297
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.240028823412211e-05,
+      "loss": 14.5527,
+      "step": 7298
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.239790676533941e-05,
+      "loss": 16.8614,
+      "step": 7299
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.239552495418228e-05,
+      "loss": 17.6649,
+      "step": 7300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.239314280066998e-05,
+      "loss": 18.5253,
+      "step": 7301
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.239076030482172e-05,
+      "loss": 16.1534,
+      "step": 7302
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.238837746665675e-05,
+      "loss": 16.8492,
+      "step": 7303
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.23859942861943e-05,
+      "loss": 15.7896,
+      "step": 7304
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.238361076345363e-05,
+      "loss": 15.3339,
+      "step": 7305
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.238122689845399e-05,
+      "loss": 16.8108,
+      "step": 7306
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.237884269121461e-05,
+      "loss": 16.7924,
+      "step": 7307
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.237645814175476e-05,
+      "loss": 16.9992,
+      "step": 7308
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.237407325009369e-05,
+      "loss": 18.049,
+      "step": 7309
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.237168801625065e-05,
+      "loss": 16.2689,
+      "step": 7310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.236930244024491e-05,
+      "loss": 16.1878,
+      "step": 7311
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.236691652209574e-05,
+      "loss": 16.8712,
+      "step": 7312
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.236453026182242e-05,
+      "loss": 17.4251,
+      "step": 7313
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.236214365944418e-05,
+      "loss": 16.6539,
+      "step": 7314
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.235975671498031e-05,
+      "loss": 16.8567,
+      "step": 7315
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.235736942845011e-05,
+      "loss": 17.9655,
+      "step": 7316
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.235498179987284e-05,
+      "loss": 16.7302,
+      "step": 7317
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.235259382926775e-05,
+      "loss": 17.098,
+      "step": 7318
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.235020551665417e-05,
+      "loss": 16.1302,
+      "step": 7319
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.234781686205138e-05,
+      "loss": 16.5715,
+      "step": 7320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.234542786547864e-05,
+      "loss": 16.7509,
+      "step": 7321
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.234303852695525e-05,
+      "loss": 16.2,
+      "step": 7322
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.234064884650054e-05,
+      "loss": 17.4413,
+      "step": 7323
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.233825882413377e-05,
+      "loss": 16.5114,
+      "step": 7324
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.233586845987424e-05,
+      "loss": 18.0078,
+      "step": 7325
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.233347775374126e-05,
+      "loss": 15.0076,
+      "step": 7326
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.233108670575414e-05,
+      "loss": 14.5732,
+      "step": 7327
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.23286953159322e-05,
+      "loss": 15.0752,
+      "step": 7328
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.232630358429474e-05,
+      "loss": 17.0582,
+      "step": 7329
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.232391151086106e-05,
+      "loss": 17.5455,
+      "step": 7330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.23215190956505e-05,
+      "loss": 17.0182,
+      "step": 7331
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.231912633868236e-05,
+      "loss": 15.9385,
+      "step": 7332
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.231673323997598e-05,
+      "loss": 16.409,
+      "step": 7333
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.231433979955066e-05,
+      "loss": 16.1312,
+      "step": 7334
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.231194601742576e-05,
+      "loss": 14.9837,
+      "step": 7335
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.230955189362058e-05,
+      "loss": 15.1747,
+      "step": 7336
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.230715742815448e-05,
+      "loss": 17.6237,
+      "step": 7337
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.230476262104677e-05,
+      "loss": 16.3323,
+      "step": 7338
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.230236747231682e-05,
+      "loss": 16.5667,
+      "step": 7339
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.229997198198393e-05,
+      "loss": 19.1206,
+      "step": 7340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.229757615006749e-05,
+      "loss": 15.4739,
+      "step": 7341
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.229517997658681e-05,
+      "loss": 16.3433,
+      "step": 7342
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.229278346156126e-05,
+      "loss": 18.0776,
+      "step": 7343
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.229038660501017e-05,
+      "loss": 15.502,
+      "step": 7344
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.228798940695293e-05,
+      "loss": 15.3536,
+      "step": 7345
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.228559186740886e-05,
+      "loss": 17.4285,
+      "step": 7346
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.228319398639735e-05,
+      "loss": 15.5762,
+      "step": 7347
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.228079576393777e-05,
+      "loss": 16.8277,
+      "step": 7348
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.227839720004945e-05,
+      "loss": 16.9742,
+      "step": 7349
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.227599829475179e-05,
+      "loss": 17.0265,
+      "step": 7350
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.227359904806414e-05,
+      "loss": 14.8754,
+      "step": 7351
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.22711994600059e-05,
+      "loss": 16.1594,
+      "step": 7352
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.226879953059641e-05,
+      "loss": 16.8062,
+      "step": 7353
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.226639925985509e-05,
+      "loss": 18.2264,
+      "step": 7354
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.22639986478013e-05,
+      "loss": 14.4259,
+      "step": 7355
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.226159769445445e-05,
+      "loss": 16.9799,
+      "step": 7356
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.22591963998339e-05,
+      "loss": 15.884,
+      "step": 7357
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.225679476395904e-05,
+      "loss": 16.4955,
+      "step": 7358
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.225439278684928e-05,
+      "loss": 16.7993,
+      "step": 7359
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.225199046852402e-05,
+      "loss": 15.4321,
+      "step": 7360
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.224958780900263e-05,
+      "loss": 18.3334,
+      "step": 7361
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.224718480830454e-05,
+      "loss": 17.4943,
+      "step": 7362
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.224478146644917e-05,
+      "loss": 18.1855,
+      "step": 7363
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.224237778345587e-05,
+      "loss": 16.614,
+      "step": 7364
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.223997375934412e-05,
+      "loss": 15.7814,
+      "step": 7365
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.223756939413329e-05,
+      "loss": 15.2535,
+      "step": 7366
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.223516468784279e-05,
+      "loss": 16.571,
+      "step": 7367
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.223275964049205e-05,
+      "loss": 15.1828,
+      "step": 7368
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.223035425210052e-05,
+      "loss": 16.9684,
+      "step": 7369
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.222794852268758e-05,
+      "loss": 16.5544,
+      "step": 7370
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.222554245227268e-05,
+      "loss": 14.4376,
+      "step": 7371
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.222313604087525e-05,
+      "loss": 17.6155,
+      "step": 7372
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.22207292885147e-05,
+      "loss": 16.9637,
+      "step": 7373
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.221832219521051e-05,
+      "loss": 15.1938,
+      "step": 7374
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.221591476098207e-05,
+      "loss": 15.8221,
+      "step": 7375
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.221350698584886e-05,
+      "loss": 17.3612,
+      "step": 7376
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.221109886983028e-05,
+      "loss": 17.3473,
+      "step": 7377
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.220869041294582e-05,
+      "loss": 14.7215,
+      "step": 7378
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.220628161521489e-05,
+      "loss": 15.9099,
+      "step": 7379
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.220387247665699e-05,
+      "loss": 16.2487,
+      "step": 7380
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.220146299729153e-05,
+      "loss": 16.6367,
+      "step": 7381
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.219905317713799e-05,
+      "loss": 16.5447,
+      "step": 7382
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.219664301621582e-05,
+      "loss": 15.184,
+      "step": 7383
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.219423251454448e-05,
+      "loss": 13.8162,
+      "step": 7384
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.219182167214344e-05,
+      "loss": 18.0614,
+      "step": 7385
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.218941048903218e-05,
+      "loss": 15.1899,
+      "step": 7386
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.218699896523016e-05,
+      "loss": 15.6305,
+      "step": 7387
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.218458710075686e-05,
+      "loss": 15.12,
+      "step": 7388
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.218217489563173e-05,
+      "loss": 16.8955,
+      "step": 7389
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.217976234987428e-05,
+      "loss": 14.0325,
+      "step": 7390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.2177349463504e-05,
+      "loss": 14.9136,
+      "step": 7391
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.217493623654035e-05,
+      "loss": 14.129,
+      "step": 7392
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.21725226690028e-05,
+      "loss": 17.023,
+      "step": 7393
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.217010876091087e-05,
+      "loss": 15.6555,
+      "step": 7394
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.216769451228407e-05,
+      "loss": 14.9282,
+      "step": 7395
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.216527992314186e-05,
+      "loss": 17.6254,
+      "step": 7396
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.216286499350375e-05,
+      "loss": 16.5707,
+      "step": 7397
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.216044972338924e-05,
+      "loss": 18.3926,
+      "step": 7398
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.215803411281784e-05,
+      "loss": 15.6286,
+      "step": 7399
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.215561816180907e-05,
+      "loss": 16.8168,
+      "step": 7400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.215320187038239e-05,
+      "loss": 19.188,
+      "step": 7401
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.215078523855737e-05,
+      "loss": 16.9341,
+      "step": 7402
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.214836826635349e-05,
+      "loss": 18.9196,
+      "step": 7403
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.214595095379029e-05,
+      "loss": 15.5252,
+      "step": 7404
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.214353330088726e-05,
+      "loss": 17.411,
+      "step": 7405
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.214111530766394e-05,
+      "loss": 16.3884,
+      "step": 7406
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.213869697413986e-05,
+      "loss": 17.3192,
+      "step": 7407
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.213627830033456e-05,
+      "loss": 17.1676,
+      "step": 7408
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.213385928626755e-05,
+      "loss": 16.951,
+      "step": 7409
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.213143993195838e-05,
+      "loss": 17.6154,
+      "step": 7410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.212902023742656e-05,
+      "loss": 16.4066,
+      "step": 7411
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.212660020269166e-05,
+      "loss": 18.0701,
+      "step": 7412
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.212417982777321e-05,
+      "loss": 18.2592,
+      "step": 7413
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.212175911269075e-05,
+      "loss": 15.3721,
+      "step": 7414
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.211933805746385e-05,
+      "loss": 14.0194,
+      "step": 7415
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.211691666211202e-05,
+      "loss": 15.5458,
+      "step": 7416
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.211449492665485e-05,
+      "loss": 17.8936,
+      "step": 7417
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.211207285111188e-05,
+      "loss": 16.3659,
+      "step": 7418
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.210965043550268e-05,
+      "loss": 14.9835,
+      "step": 7419
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.210722767984678e-05,
+      "loss": 16.7887,
+      "step": 7420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.210480458416379e-05,
+      "loss": 14.965,
+      "step": 7421
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.210238114847328e-05,
+      "loss": 16.9167,
+      "step": 7422
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.209995737279476e-05,
+      "loss": 15.2917,
+      "step": 7423
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.209753325714785e-05,
+      "loss": 14.7762,
+      "step": 7424
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.20951088015521e-05,
+      "loss": 15.3659,
+      "step": 7425
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.209268400602712e-05,
+      "loss": 15.7514,
+      "step": 7426
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.209025887059246e-05,
+      "loss": 16.8193,
+      "step": 7427
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.208783339526773e-05,
+      "loss": 17.9097,
+      "step": 7428
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.208540758007249e-05,
+      "loss": 16.2016,
+      "step": 7429
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.208298142502636e-05,
+      "loss": 17.7816,
+      "step": 7430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.20805549301489e-05,
+      "loss": 15.9132,
+      "step": 7431
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.207812809545973e-05,
+      "loss": 15.5172,
+      "step": 7432
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.207570092097842e-05,
+      "loss": 16.6549,
+      "step": 7433
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.207327340672459e-05,
+      "loss": 15.0805,
+      "step": 7434
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.207084555271784e-05,
+      "loss": 16.9535,
+      "step": 7435
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.206841735897778e-05,
+      "loss": 15.9638,
+      "step": 7436
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.206598882552399e-05,
+      "loss": 14.6324,
+      "step": 7437
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.206355995237614e-05,
+      "loss": 14.4052,
+      "step": 7438
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.206113073955378e-05,
+      "loss": 13.9679,
+      "step": 7439
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.205870118707657e-05,
+      "loss": 14.9878,
+      "step": 7440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.20562712949641e-05,
+      "loss": 15.1552,
+      "step": 7441
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.205384106323602e-05,
+      "loss": 15.6217,
+      "step": 7442
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.205141049191192e-05,
+      "loss": 17.9146,
+      "step": 7443
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.204897958101146e-05,
+      "loss": 16.3327,
+      "step": 7444
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.204654833055427e-05,
+      "loss": 14.8229,
+      "step": 7445
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.204411674055994e-05,
+      "loss": 15.3369,
+      "step": 7446
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.204168481104815e-05,
+      "loss": 16.9417,
+      "step": 7447
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.203925254203853e-05,
+      "loss": 16.8002,
+      "step": 7448
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.203681993355071e-05,
+      "loss": 16.873,
+      "step": 7449
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.203438698560436e-05,
+      "loss": 16.9874,
+      "step": 7450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.203195369821908e-05,
+      "loss": 16.5867,
+      "step": 7451
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.202952007141456e-05,
+      "loss": 18.051,
+      "step": 7452
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.202708610521045e-05,
+      "loss": 15.633,
+      "step": 7453
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.202465179962637e-05,
+      "loss": 16.4584,
+      "step": 7454
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.202221715468201e-05,
+      "loss": 17.9519,
+      "step": 7455
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.201978217039702e-05,
+      "loss": 15.5135,
+      "step": 7456
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.201734684679107e-05,
+      "loss": 15.1176,
+      "step": 7457
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.201491118388381e-05,
+      "loss": 16.9959,
+      "step": 7458
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.201247518169494e-05,
+      "loss": 17.0005,
+      "step": 7459
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.20100388402441e-05,
+      "loss": 15.6433,
+      "step": 7460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.200760215955097e-05,
+      "loss": 16.3686,
+      "step": 7461
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.200516513963523e-05,
+      "loss": 16.988,
+      "step": 7462
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.200272778051656e-05,
+      "loss": 17.0709,
+      "step": 7463
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.200029008221465e-05,
+      "loss": 16.9595,
+      "step": 7464
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.199785204474918e-05,
+      "loss": 16.4298,
+      "step": 7465
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.199541366813982e-05,
+      "loss": 17.2022,
+      "step": 7466
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.19929749524063e-05,
+      "loss": 16.6932,
+      "step": 7467
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.199053589756827e-05,
+      "loss": 17.7191,
+      "step": 7468
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.198809650364546e-05,
+      "loss": 16.561,
+      "step": 7469
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.198565677065756e-05,
+      "loss": 14.796,
+      "step": 7470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.198321669862426e-05,
+      "loss": 17.5588,
+      "step": 7471
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.198077628756525e-05,
+      "loss": 17.3329,
+      "step": 7472
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.197833553750028e-05,
+      "loss": 17.335,
+      "step": 7473
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.197589444844904e-05,
+      "loss": 17.4609,
+      "step": 7474
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.197345302043124e-05,
+      "loss": 16.5342,
+      "step": 7475
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.197101125346657e-05,
+      "loss": 18.0313,
+      "step": 7476
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.196856914757478e-05,
+      "loss": 15.9255,
+      "step": 7477
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.196612670277561e-05,
+      "loss": 16.1716,
+      "step": 7478
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.196368391908873e-05,
+      "loss": 15.1841,
+      "step": 7479
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.196124079653393e-05,
+      "loss": 16.9811,
+      "step": 7480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.195879733513087e-05,
+      "loss": 14.9452,
+      "step": 7481
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.195635353489931e-05,
+      "loss": 17.9648,
+      "step": 7482
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.1953909395859e-05,
+      "loss": 17.7928,
+      "step": 7483
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.195146491802967e-05,
+      "loss": 17.3746,
+      "step": 7484
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.194902010143103e-05,
+      "loss": 16.2081,
+      "step": 7485
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.194657494608287e-05,
+      "loss": 15.8989,
+      "step": 7486
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.19441294520049e-05,
+      "loss": 17.595,
+      "step": 7487
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.194168361921689e-05,
+      "loss": 16.0499,
+      "step": 7488
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.193923744773858e-05,
+      "loss": 16.3804,
+      "step": 7489
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.193679093758971e-05,
+      "loss": 17.6254,
+      "step": 7490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.193434408879007e-05,
+      "loss": 16.6435,
+      "step": 7491
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.19318969013594e-05,
+      "loss": 15.6089,
+      "step": 7492
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.192944937531744e-05,
+      "loss": 15.5705,
+      "step": 7493
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.1927001510684e-05,
+      "loss": 16.3384,
+      "step": 7494
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.192455330747881e-05,
+      "loss": 16.1642,
+      "step": 7495
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.192210476572166e-05,
+      "loss": 15.8055,
+      "step": 7496
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.191965588543232e-05,
+      "loss": 16.4869,
+      "step": 7497
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.191720666663055e-05,
+      "loss": 15.3296,
+      "step": 7498
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.191475710933615e-05,
+      "loss": 14.0477,
+      "step": 7499
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.191230721356888e-05,
+      "loss": 15.433,
+      "step": 7500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.190985697934855e-05,
+      "loss": 14.9708,
+      "step": 7501
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.19074064066949e-05,
+      "loss": 16.7374,
+      "step": 7502
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.190495549562776e-05,
+      "loss": 15.608,
+      "step": 7503
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.190250424616693e-05,
+      "loss": 16.9509,
+      "step": 7504
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.190005265833217e-05,
+      "loss": 15.1489,
+      "step": 7505
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.18976007321433e-05,
+      "loss": 15.0596,
+      "step": 7506
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.18951484676201e-05,
+      "loss": 17.2253,
+      "step": 7507
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.18926958647824e-05,
+      "loss": 17.1032,
+      "step": 7508
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.189024292364999e-05,
+      "loss": 16.1882,
+      "step": 7509
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.188778964424266e-05,
+      "loss": 17.3169,
+      "step": 7510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.188533602658025e-05,
+      "loss": 17.0831,
+      "step": 7511
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.188288207068257e-05,
+      "loss": 19.0838,
+      "step": 7512
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.188042777656942e-05,
+      "loss": 16.3842,
+      "step": 7513
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.187797314426064e-05,
+      "loss": 14.0825,
+      "step": 7514
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.187551817377602e-05,
+      "loss": 17.4036,
+      "step": 7515
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.187306286513541e-05,
+      "loss": 16.567,
+      "step": 7516
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.187060721835866e-05,
+      "loss": 16.0217,
+      "step": 7517
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.186815123346555e-05,
+      "loss": 14.7111,
+      "step": 7518
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.186569491047593e-05,
+      "loss": 15.5415,
+      "step": 7519
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.186323824940966e-05,
+      "loss": 15.376,
+      "step": 7520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.186078125028654e-05,
+      "loss": 16.2663,
+      "step": 7521
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.185832391312644e-05,
+      "loss": 15.3129,
+      "step": 7522
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.185586623794919e-05,
+      "loss": 17.4424,
+      "step": 7523
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.185340822477464e-05,
+      "loss": 17.0127,
+      "step": 7524
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.185094987362264e-05,
+      "loss": 18.2125,
+      "step": 7525
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.184849118451303e-05,
+      "loss": 17.6642,
+      "step": 7526
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.184603215746568e-05,
+      "loss": 15.2689,
+      "step": 7527
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.184357279250044e-05,
+      "loss": 15.1806,
+      "step": 7528
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.184111308963719e-05,
+      "loss": 17.7437,
+      "step": 7529
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.183865304889576e-05,
+      "loss": 15.0747,
+      "step": 7530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.183619267029604e-05,
+      "loss": 16.0844,
+      "step": 7531
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.183373195385788e-05,
+      "loss": 15.677,
+      "step": 7532
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.183127089960115e-05,
+      "loss": 15.7906,
+      "step": 7533
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.182880950754576e-05,
+      "loss": 17.5072,
+      "step": 7534
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.182634777771155e-05,
+      "loss": 15.9474,
+      "step": 7535
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.18238857101184e-05,
+      "loss": 15.003,
+      "step": 7536
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.18214233047862e-05,
+      "loss": 15.528,
+      "step": 7537
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.181896056173484e-05,
+      "loss": 15.8594,
+      "step": 7538
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.18164974809842e-05,
+      "loss": 18.7414,
+      "step": 7539
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.181403406255418e-05,
+      "loss": 15.0767,
+      "step": 7540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.181157030646467e-05,
+      "loss": 15.885,
+      "step": 7541
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.180910621273555e-05,
+      "loss": 16.7601,
+      "step": 7542
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.180664178138674e-05,
+      "loss": 15.873,
+      "step": 7543
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.18041770124381e-05,
+      "loss": 14.7989,
+      "step": 7544
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.18017119059096e-05,
+      "loss": 17.1397,
+      "step": 7545
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.179924646182109e-05,
+      "loss": 15.3917,
+      "step": 7546
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.179678068019252e-05,
+      "loss": 16.1509,
+      "step": 7547
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.179431456104377e-05,
+      "loss": 16.5542,
+      "step": 7548
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.179184810439477e-05,
+      "loss": 16.8568,
+      "step": 7549
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.178938131026542e-05,
+      "loss": 18.3694,
+      "step": 7550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.178691417867565e-05,
+      "loss": 16.9672,
+      "step": 7551
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.17844467096454e-05,
+      "loss": 17.7404,
+      "step": 7552
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.178197890319458e-05,
+      "loss": 17.1047,
+      "step": 7553
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.177951075934312e-05,
+      "loss": 16.3262,
+      "step": 7554
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.177704227811094e-05,
+      "loss": 16.2391,
+      "step": 7555
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.177457345951798e-05,
+      "loss": 15.9639,
+      "step": 7556
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.17721043035842e-05,
+      "loss": 16.5525,
+      "step": 7557
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.176963481032951e-05,
+      "loss": 16.7464,
+      "step": 7558
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.176716497977387e-05,
+      "loss": 16.3941,
+      "step": 7559
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.176469481193722e-05,
+      "loss": 15.3191,
+      "step": 7560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.176222430683947e-05,
+      "loss": 14.8903,
+      "step": 7561
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.175975346450063e-05,
+      "loss": 17.8696,
+      "step": 7562
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.175728228494063e-05,
+      "loss": 16.2657,
+      "step": 7563
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.175481076817942e-05,
+      "loss": 19.2643,
+      "step": 7564
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.175233891423695e-05,
+      "loss": 16.974,
+      "step": 7565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.17498667231332e-05,
+      "loss": 16.0355,
+      "step": 7566
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.17473941948881e-05,
+      "loss": 16.7613,
+      "step": 7567
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.174492132952166e-05,
+      "loss": 17.0694,
+      "step": 7568
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.174244812705382e-05,
+      "loss": 17.3965,
+      "step": 7569
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.173997458750457e-05,
+      "loss": 17.2598,
+      "step": 7570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.173750071089388e-05,
+      "loss": 16.654,
+      "step": 7571
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.17350264972417e-05,
+      "loss": 15.2003,
+      "step": 7572
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.173255194656806e-05,
+      "loss": 16.4306,
+      "step": 7573
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.173007705889291e-05,
+      "loss": 17.4111,
+      "step": 7574
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.172760183423623e-05,
+      "loss": 18.0296,
+      "step": 7575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.172512627261802e-05,
+      "loss": 16.0854,
+      "step": 7576
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.172265037405827e-05,
+      "loss": 18.2595,
+      "step": 7577
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.172017413857698e-05,
+      "loss": 14.9559,
+      "step": 7578
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.171769756619414e-05,
+      "loss": 18.5378,
+      "step": 7579
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.171522065692975e-05,
+      "loss": 16.0936,
+      "step": 7580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.17127434108038e-05,
+      "loss": 16.575,
+      "step": 7581
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.171026582783631e-05,
+      "loss": 16.5608,
+      "step": 7582
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.170778790804729e-05,
+      "loss": 15.6375,
+      "step": 7583
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.170530965145671e-05,
+      "loss": 15.8615,
+      "step": 7584
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.170283105808466e-05,
+      "loss": 15.7115,
+      "step": 7585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.170035212795107e-05,
+      "loss": 17.3074,
+      "step": 7586
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.169787286107603e-05,
+      "loss": 15.4024,
+      "step": 7587
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.169539325747951e-05,
+      "loss": 16.6741,
+      "step": 7588
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.169291331718156e-05,
+      "loss": 14.9879,
+      "step": 7589
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.169043304020218e-05,
+      "loss": 17.5565,
+      "step": 7590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.168795242656142e-05,
+      "loss": 15.2283,
+      "step": 7591
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.168547147627932e-05,
+      "loss": 15.8863,
+      "step": 7592
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.168299018937592e-05,
+      "loss": 16.0471,
+      "step": 7593
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.168050856587121e-05,
+      "loss": 17.8764,
+      "step": 7594
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.167802660578527e-05,
+      "loss": 14.8817,
+      "step": 7595
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.167554430913815e-05,
+      "loss": 17.1147,
+      "step": 7596
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.167306167594984e-05,
+      "loss": 17.4558,
+      "step": 7597
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.167057870624045e-05,
+      "loss": 16.7197,
+      "step": 7598
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.166809540003e-05,
+      "loss": 15.1259,
+      "step": 7599
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.166561175733854e-05,
+      "loss": 16.6931,
+      "step": 7600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.166312777818615e-05,
+      "loss": 16.7052,
+      "step": 7601
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.166064346259287e-05,
+      "loss": 14.5582,
+      "step": 7602
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.165815881057878e-05,
+      "loss": 16.0507,
+      "step": 7603
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.165567382216391e-05,
+      "loss": 16.3757,
+      "step": 7604
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.165318849736835e-05,
+      "loss": 17.4164,
+      "step": 7605
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.165070283621217e-05,
+      "loss": 15.1247,
+      "step": 7606
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.164821683871543e-05,
+      "loss": 15.5254,
+      "step": 7607
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.164573050489822e-05,
+      "loss": 16.5321,
+      "step": 7608
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.164324383478062e-05,
+      "loss": 14.0783,
+      "step": 7609
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.164075682838269e-05,
+      "loss": 16.0016,
+      "step": 7610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.163826948572454e-05,
+      "loss": 17.5198,
+      "step": 7611
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.163578180682623e-05,
+      "loss": 16.3945,
+      "step": 7612
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.163329379170786e-05,
+      "loss": 17.8707,
+      "step": 7613
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.163080544038952e-05,
+      "loss": 18.2373,
+      "step": 7614
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.162831675289132e-05,
+      "loss": 16.4737,
+      "step": 7615
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.162582772923332e-05,
+      "loss": 15.8147,
+      "step": 7616
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.162333836943566e-05,
+      "loss": 16.5274,
+      "step": 7617
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.162084867351842e-05,
+      "loss": 15.8569,
+      "step": 7618
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.16183586415017e-05,
+      "loss": 15.9677,
+      "step": 7619
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.161586827340563e-05,
+      "loss": 14.5986,
+      "step": 7620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.16133775692503e-05,
+      "loss": 17.347,
+      "step": 7621
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.161088652905583e-05,
+      "loss": 15.1888,
+      "step": 7622
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.160839515284234e-05,
+      "loss": 14.3789,
+      "step": 7623
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.160590344062994e-05,
+      "loss": 14.6758,
+      "step": 7624
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.160341139243875e-05,
+      "loss": 18.1841,
+      "step": 7625
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.160091900828892e-05,
+      "loss": 14.5301,
+      "step": 7626
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.159842628820054e-05,
+      "loss": 14.6892,
+      "step": 7627
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.159593323219376e-05,
+      "loss": 15.3781,
+      "step": 7628
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.159343984028872e-05,
+      "loss": 16.1738,
+      "step": 7629
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.159094611250553e-05,
+      "loss": 15.1859,
+      "step": 7630
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.158845204886434e-05,
+      "loss": 18.6668,
+      "step": 7631
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.158595764938529e-05,
+      "loss": 16.0309,
+      "step": 7632
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.158346291408852e-05,
+      "loss": 14.9391,
+      "step": 7633
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.15809678429942e-05,
+      "loss": 15.6574,
+      "step": 7634
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.157847243612245e-05,
+      "loss": 15.4502,
+      "step": 7635
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.157597669349342e-05,
+      "loss": 17.6276,
+      "step": 7636
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.157348061512727e-05,
+      "loss": 16.6996,
+      "step": 7637
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.157098420104416e-05,
+      "loss": 16.4866,
+      "step": 7638
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.156848745126425e-05,
+      "loss": 15.9391,
+      "step": 7639
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.156599036580771e-05,
+      "loss": 16.6699,
+      "step": 7640
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.156349294469468e-05,
+      "loss": 16.2198,
+      "step": 7641
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.156099518794534e-05,
+      "loss": 17.3122,
+      "step": 7642
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.155849709557989e-05,
+      "loss": 16.4446,
+      "step": 7643
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.155599866761844e-05,
+      "loss": 16.2113,
+      "step": 7644
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.155349990408122e-05,
+      "loss": 15.4548,
+      "step": 7645
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.155100080498839e-05,
+      "loss": 18.3303,
+      "step": 7646
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.154850137036012e-05,
+      "loss": 16.0218,
+      "step": 7647
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.15460016002166e-05,
+      "loss": 17.5898,
+      "step": 7648
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.154350149457803e-05,
+      "loss": 15.5198,
+      "step": 7649
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.154100105346459e-05,
+      "loss": 15.3922,
+      "step": 7650
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.153850027689645e-05,
+      "loss": 15.5912,
+      "step": 7651
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.153599916489384e-05,
+      "loss": 15.1084,
+      "step": 7652
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.153349771747694e-05,
+      "loss": 16.6227,
+      "step": 7653
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.153099593466597e-05,
+      "loss": 15.5055,
+      "step": 7654
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.152849381648108e-05,
+      "loss": 15.6809,
+      "step": 7655
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.152599136294253e-05,
+      "loss": 17.5596,
+      "step": 7656
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.152348857407051e-05,
+      "loss": 17.1847,
+      "step": 7657
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.15209854498852e-05,
+      "loss": 15.6446,
+      "step": 7658
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.151848199040687e-05,
+      "loss": 15.167,
+      "step": 7659
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.151597819565571e-05,
+      "loss": 14.3744,
+      "step": 7660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.151347406565192e-05,
+      "loss": 16.4903,
+      "step": 7661
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.151096960041574e-05,
+      "loss": 15.9275,
+      "step": 7662
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.15084647999674e-05,
+      "loss": 15.8552,
+      "step": 7663
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.150595966432712e-05,
+      "loss": 15.9243,
+      "step": 7664
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.150345419351511e-05,
+      "loss": 17.4983,
+      "step": 7665
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.150094838755164e-05,
+      "loss": 17.7656,
+      "step": 7666
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.149844224645691e-05,
+      "loss": 18.4362,
+      "step": 7667
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.14959357702512e-05,
+      "loss": 16.4804,
+      "step": 7668
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.149342895895471e-05,
+      "loss": 17.1221,
+      "step": 7669
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.14909218125877e-05,
+      "loss": 14.9433,
+      "step": 7670
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.148841433117041e-05,
+      "loss": 15.0647,
+      "step": 7671
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.148590651472309e-05,
+      "loss": 18.7592,
+      "step": 7672
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.148339836326598e-05,
+      "loss": 15.388,
+      "step": 7673
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.148088987681938e-05,
+      "loss": 18.3208,
+      "step": 7674
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.147838105540348e-05,
+      "loss": 16.6121,
+      "step": 7675
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.147587189903859e-05,
+      "loss": 18.0049,
+      "step": 7676
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.147336240774495e-05,
+      "loss": 15.0399,
+      "step": 7677
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.147085258154284e-05,
+      "loss": 18.488,
+      "step": 7678
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.146834242045252e-05,
+      "loss": 16.6307,
+      "step": 7679
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.146583192449425e-05,
+      "loss": 15.036,
+      "step": 7680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.14633210936883e-05,
+      "loss": 15.5505,
+      "step": 7681
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.146080992805497e-05,
+      "loss": 17.8134,
+      "step": 7682
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.145829842761453e-05,
+      "loss": 16.8507,
+      "step": 7683
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.145578659238725e-05,
+      "loss": 15.423,
+      "step": 7684
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.145327442239341e-05,
+      "loss": 16.2077,
+      "step": 7685
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.145076191765332e-05,
+      "loss": 16.0013,
+      "step": 7686
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.144824907818724e-05,
+      "loss": 17.9065,
+      "step": 7687
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.144573590401549e-05,
+      "loss": 15.7058,
+      "step": 7688
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.144322239515834e-05,
+      "loss": 14.9394,
+      "step": 7689
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.144070855163611e-05,
+      "loss": 16.7288,
+      "step": 7690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.143819437346909e-05,
+      "loss": 18.1804,
+      "step": 7691
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.143567986067757e-05,
+      "loss": 15.0376,
+      "step": 7692
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.143316501328188e-05,
+      "loss": 14.4607,
+      "step": 7693
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.14306498313023e-05,
+      "loss": 16.816,
+      "step": 7694
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.142813431475916e-05,
+      "loss": 17.0851,
+      "step": 7695
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.142561846367278e-05,
+      "loss": 16.554,
+      "step": 7696
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.142310227806346e-05,
+      "loss": 14.479,
+      "step": 7697
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.142058575795152e-05,
+      "loss": 15.2831,
+      "step": 7698
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.141806890335729e-05,
+      "loss": 15.7781,
+      "step": 7699
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.14155517143011e-05,
+      "loss": 16.0049,
+      "step": 7700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.141303419080325e-05,
+      "loss": 17.4209,
+      "step": 7701
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.141051633288409e-05,
+      "loss": 17.152,
+      "step": 7702
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.140799814056395e-05,
+      "loss": 16.6472,
+      "step": 7703
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.140547961386316e-05,
+      "loss": 15.4401,
+      "step": 7704
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.140296075280207e-05,
+      "loss": 15.9624,
+      "step": 7705
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.140044155740101e-05,
+      "loss": 16.054,
+      "step": 7706
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.139792202768031e-05,
+      "loss": 16.1772,
+      "step": 7707
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.139540216366035e-05,
+      "loss": 16.458,
+      "step": 7708
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.139288196536147e-05,
+      "loss": 15.915,
+      "step": 7709
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.139036143280399e-05,
+      "loss": 15.743,
+      "step": 7710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.138784056600829e-05,
+      "loss": 15.286,
+      "step": 7711
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.138531936499473e-05,
+      "loss": 16.462,
+      "step": 7712
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.138279782978366e-05,
+      "loss": 15.8079,
+      "step": 7713
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.138027596039544e-05,
+      "loss": 15.5521,
+      "step": 7714
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.137775375685042e-05,
+      "loss": 16.6934,
+      "step": 7715
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.137523121916903e-05,
+      "loss": 14.9781,
+      "step": 7716
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.137270834737155e-05,
+      "loss": 17.391,
+      "step": 7717
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.137018514147842e-05,
+      "loss": 17.2935,
+      "step": 7718
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.136766160151e-05,
+      "loss": 14.9034,
+      "step": 7719
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.136513772748665e-05,
+      "loss": 15.8172,
+      "step": 7720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.136261351942877e-05,
+      "loss": 13.7827,
+      "step": 7721
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.136008897735673e-05,
+      "loss": 15.2171,
+      "step": 7722
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.135756410129093e-05,
+      "loss": 15.3026,
+      "step": 7723
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.135503889125175e-05,
+      "loss": 16.4104,
+      "step": 7724
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.135251334725959e-05,
+      "loss": 14.7194,
+      "step": 7725
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.134998746933483e-05,
+      "loss": 18.6816,
+      "step": 7726
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.134746125749787e-05,
+      "loss": 19.612,
+      "step": 7727
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.134493471176912e-05,
+      "loss": 17.7072,
+      "step": 7728
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.134240783216899e-05,
+      "loss": 16.2699,
+      "step": 7729
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.133988061871787e-05,
+      "loss": 15.9675,
+      "step": 7730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.133735307143617e-05,
+      "loss": 14.8571,
+      "step": 7731
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.133482519034428e-05,
+      "loss": 16.5441,
+      "step": 7732
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.133229697546267e-05,
+      "loss": 17.5997,
+      "step": 7733
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.132976842681172e-05,
+      "loss": 17.0374,
+      "step": 7734
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.132723954441183e-05,
+      "loss": 16.3129,
+      "step": 7735
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.132471032828344e-05,
+      "loss": 16.8311,
+      "step": 7736
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.132218077844697e-05,
+      "loss": 14.464,
+      "step": 7737
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.131965089492286e-05,
+      "loss": 16.5518,
+      "step": 7738
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.131712067773155e-05,
+      "loss": 15.8378,
+      "step": 7739
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.131459012689344e-05,
+      "loss": 16.5688,
+      "step": 7740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.131205924242898e-05,
+      "loss": 16.2511,
+      "step": 7741
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.13095280243586e-05,
+      "loss": 15.1677,
+      "step": 7742
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.130699647270276e-05,
+      "loss": 17.4979,
+      "step": 7743
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.130446458748188e-05,
+      "loss": 16.3874,
+      "step": 7744
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.130193236871641e-05,
+      "loss": 15.9485,
+      "step": 7745
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.129939981642681e-05,
+      "loss": 18.3872,
+      "step": 7746
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.129686693063352e-05,
+      "loss": 17.8688,
+      "step": 7747
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.129433371135701e-05,
+      "loss": 16.7181,
+      "step": 7748
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.129180015861771e-05,
+      "loss": 16.4919,
+      "step": 7749
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.128926627243612e-05,
+      "loss": 14.5273,
+      "step": 7750
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.128673205283265e-05,
+      "loss": 16.4544,
+      "step": 7751
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.12841974998278e-05,
+      "loss": 17.4782,
+      "step": 7752
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.128166261344202e-05,
+      "loss": 17.1697,
+      "step": 7753
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.127912739369579e-05,
+      "loss": 15.4293,
+      "step": 7754
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.12765918406096e-05,
+      "loss": 16.0464,
+      "step": 7755
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.127405595420387e-05,
+      "loss": 15.802,
+      "step": 7756
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.127151973449913e-05,
+      "loss": 14.6671,
+      "step": 7757
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.126898318151585e-05,
+      "loss": 15.485,
+      "step": 7758
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.12664462952745e-05,
+      "loss": 17.3377,
+      "step": 7759
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.126390907579557e-05,
+      "loss": 16.5197,
+      "step": 7760
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.126137152309956e-05,
+      "loss": 16.0444,
+      "step": 7761
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.125883363720695e-05,
+      "loss": 15.7506,
+      "step": 7762
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.125629541813823e-05,
+      "loss": 15.8354,
+      "step": 7763
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.125375686591391e-05,
+      "loss": 16.019,
+      "step": 7764
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.125121798055449e-05,
+      "loss": 14.6331,
+      "step": 7765
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.124867876208045e-05,
+      "loss": 16.8404,
+      "step": 7766
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.124613921051233e-05,
+      "loss": 16.4761,
+      "step": 7767
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.124359932587061e-05,
+      "loss": 16.7661,
+      "step": 7768
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.12410591081758e-05,
+      "loss": 15.3576,
+      "step": 7769
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.123851855744843e-05,
+      "loss": 16.2972,
+      "step": 7770
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.1235977673709e-05,
+      "loss": 17.2689,
+      "step": 7771
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.123343645697805e-05,
+      "loss": 18.0851,
+      "step": 7772
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.123089490727608e-05,
+      "loss": 17.3628,
+      "step": 7773
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.122835302462361e-05,
+      "loss": 15.6211,
+      "step": 7774
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.122581080904117e-05,
+      "loss": 18.6364,
+      "step": 7775
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.122326826054932e-05,
+      "loss": 15.6442,
+      "step": 7776
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.122072537916857e-05,
+      "loss": 16.4006,
+      "step": 7777
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.121818216491944e-05,
+      "loss": 16.568,
+      "step": 7778
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.121563861782247e-05,
+      "loss": 17.3506,
+      "step": 7779
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.121309473789821e-05,
+      "loss": 16.9983,
+      "step": 7780
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.12105505251672e-05,
+      "loss": 16.3841,
+      "step": 7781
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.120800597965e-05,
+      "loss": 16.0403,
+      "step": 7782
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.120546110136714e-05,
+      "loss": 16.1871,
+      "step": 7783
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.120291589033917e-05,
+      "loss": 18.256,
+      "step": 7784
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.120037034658664e-05,
+      "loss": 15.58,
+      "step": 7785
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.119782447013014e-05,
+      "loss": 17.5007,
+      "step": 7786
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.119527826099017e-05,
+      "loss": 16.9239,
+      "step": 7787
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.119273171918735e-05,
+      "loss": 15.8135,
+      "step": 7788
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.119018484474221e-05,
+      "loss": 15.8949,
+      "step": 7789
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.118763763767533e-05,
+      "loss": 18.7848,
+      "step": 7790
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.118509009800726e-05,
+      "loss": 17.9609,
+      "step": 7791
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.118254222575859e-05,
+      "loss": 14.9005,
+      "step": 7792
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.11799940209499e-05,
+      "loss": 16.2026,
+      "step": 7793
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.117744548360175e-05,
+      "loss": 15.6028,
+      "step": 7794
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.117489661373473e-05,
+      "loss": 18.7854,
+      "step": 7795
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.117234741136942e-05,
+      "loss": 15.3941,
+      "step": 7796
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.116979787652643e-05,
+      "loss": 16.3413,
+      "step": 7797
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.116724800922629e-05,
+      "loss": 18.2602,
+      "step": 7798
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.116469780948964e-05,
+      "loss": 16.0934,
+      "step": 7799
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.116214727733706e-05,
+      "loss": 16.4782,
+      "step": 7800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.115959641278915e-05,
+      "loss": 16.3814,
+      "step": 7801
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.11570452158665e-05,
+      "loss": 18.5547,
+      "step": 7802
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.11544936865897e-05,
+      "loss": 14.9729,
+      "step": 7803
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.11519418249794e-05,
+      "loss": 16.1596,
+      "step": 7804
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.114938963105614e-05,
+      "loss": 15.1381,
+      "step": 7805
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.11468371048406e-05,
+      "loss": 15.0001,
+      "step": 7806
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.114428424635335e-05,
+      "loss": 16.4147,
+      "step": 7807
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.114173105561501e-05,
+      "loss": 15.3265,
+      "step": 7808
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.11391775326462e-05,
+      "loss": 16.3599,
+      "step": 7809
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.113662367746756e-05,
+      "loss": 16.2039,
+      "step": 7810
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.113406949009968e-05,
+      "loss": 16.8309,
+      "step": 7811
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.11315149705632e-05,
+      "loss": 18.0515,
+      "step": 7812
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.112896011887874e-05,
+      "loss": 16.9311,
+      "step": 7813
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.112640493506697e-05,
+      "loss": 16.426,
+      "step": 7814
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.112384941914848e-05,
+      "loss": 16.5513,
+      "step": 7815
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.112129357114391e-05,
+      "loss": 15.6446,
+      "step": 7816
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.111873739107393e-05,
+      "loss": 17.1335,
+      "step": 7817
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.111618087895915e-05,
+      "loss": 18.1248,
+      "step": 7818
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.111362403482022e-05,
+      "loss": 15.3102,
+      "step": 7819
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.11110668586778e-05,
+      "loss": 15.9934,
+      "step": 7820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.110850935055254e-05,
+      "loss": 15.7951,
+      "step": 7821
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.110595151046508e-05,
+      "loss": 16.1058,
+      "step": 7822
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.110339333843608e-05,
+      "loss": 15.3696,
+      "step": 7823
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.110083483448621e-05,
+      "loss": 15.0129,
+      "step": 7824
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.109827599863611e-05,
+      "loss": 13.3275,
+      "step": 7825
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.109571683090645e-05,
+      "loss": 15.0754,
+      "step": 7826
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.109315733131791e-05,
+      "loss": 18.4319,
+      "step": 7827
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.109059749989114e-05,
+      "loss": 14.1246,
+      "step": 7828
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.108803733664681e-05,
+      "loss": 15.2623,
+      "step": 7829
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.108547684160562e-05,
+      "loss": 16.0161,
+      "step": 7830
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.108291601478822e-05,
+      "loss": 16.2781,
+      "step": 7831
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.108035485621531e-05,
+      "loss": 17.2913,
+      "step": 7832
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.107779336590755e-05,
+      "loss": 16.5155,
+      "step": 7833
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.107523154388565e-05,
+      "loss": 17.2878,
+      "step": 7834
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.107266939017024e-05,
+      "loss": 17.7319,
+      "step": 7835
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.10701069047821e-05,
+      "loss": 20.0217,
+      "step": 7836
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.106754408774188e-05,
+      "loss": 17.4395,
+      "step": 7837
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.106498093907024e-05,
+      "loss": 17.7051,
+      "step": 7838
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.106241745878792e-05,
+      "loss": 17.183,
+      "step": 7839
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.105985364691562e-05,
+      "loss": 16.2489,
+      "step": 7840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.105728950347403e-05,
+      "loss": 17.7707,
+      "step": 7841
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.105472502848385e-05,
+      "loss": 17.7921,
+      "step": 7842
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.10521602219658e-05,
+      "loss": 16.7915,
+      "step": 7843
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.104959508394061e-05,
+      "loss": 17.0577,
+      "step": 7844
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.104702961442898e-05,
+      "loss": 15.6653,
+      "step": 7845
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.104446381345159e-05,
+      "loss": 15.6823,
+      "step": 7846
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.10418976810292e-05,
+      "loss": 17.344,
+      "step": 7847
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.103933121718253e-05,
+      "loss": 15.5974,
+      "step": 7848
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.10367644219323e-05,
+      "loss": 17.6551,
+      "step": 7849
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.103419729529922e-05,
+      "loss": 14.998,
+      "step": 7850
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.103162983730405e-05,
+      "loss": 16.4379,
+      "step": 7851
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.10290620479675e-05,
+      "loss": 16.4849,
+      "step": 7852
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.102649392731031e-05,
+      "loss": 16.4893,
+      "step": 7853
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.102392547535324e-05,
+      "loss": 18.6986,
+      "step": 7854
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.1021356692117e-05,
+      "loss": 17.8715,
+      "step": 7855
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.101878757762236e-05,
+      "loss": 15.6078,
+      "step": 7856
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.101621813189004e-05,
+      "loss": 16.9393,
+      "step": 7857
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.101364835494081e-05,
+      "loss": 14.9998,
+      "step": 7858
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.101107824679541e-05,
+      "loss": 15.0654,
+      "step": 7859
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.100850780747461e-05,
+      "loss": 16.2331,
+      "step": 7860
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.100593703699913e-05,
+      "loss": 16.2509,
+      "step": 7861
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.100336593538979e-05,
+      "loss": 14.5677,
+      "step": 7862
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.100079450266729e-05,
+      "loss": 14.4994,
+      "step": 7863
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.099822273885244e-05,
+      "loss": 15.6069,
+      "step": 7864
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.099565064396597e-05,
+      "loss": 15.8441,
+      "step": 7865
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.099307821802867e-05,
+      "loss": 15.5889,
+      "step": 7866
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.099050546106133e-05,
+      "loss": 15.422,
+      "step": 7867
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.098793237308469e-05,
+      "loss": 16.5065,
+      "step": 7868
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.098535895411957e-05,
+      "loss": 15.5806,
+      "step": 7869
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.098278520418672e-05,
+      "loss": 17.1466,
+      "step": 7870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.098021112330692e-05,
+      "loss": 16.4074,
+      "step": 7871
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.0977636711501e-05,
+      "loss": 17.7625,
+      "step": 7872
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.09750619687897e-05,
+      "loss": 17.7127,
+      "step": 7873
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.097248689519381e-05,
+      "loss": 18.2937,
+      "step": 7874
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.096991149073416e-05,
+      "loss": 16.8364,
+      "step": 7875
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.096733575543153e-05,
+      "loss": 15.7177,
+      "step": 7876
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.096475968930672e-05,
+      "loss": 16.0051,
+      "step": 7877
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.096218329238053e-05,
+      "loss": 16.2551,
+      "step": 7878
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.095960656467377e-05,
+      "loss": 16.5578,
+      "step": 7879
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.095702950620725e-05,
+      "loss": 16.3684,
+      "step": 7880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.095445211700178e-05,
+      "loss": 16.1108,
+      "step": 7881
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.095187439707816e-05,
+      "loss": 15.667,
+      "step": 7882
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.094929634645722e-05,
+      "loss": 17.124,
+      "step": 7883
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.094671796515978e-05,
+      "loss": 15.0098,
+      "step": 7884
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.094413925320664e-05,
+      "loss": 17.3487,
+      "step": 7885
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.094156021061865e-05,
+      "loss": 17.8071,
+      "step": 7886
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.093898083741662e-05,
+      "loss": 17.281,
+      "step": 7887
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.093640113362138e-05,
+      "loss": 16.4838,
+      "step": 7888
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.093382109925377e-05,
+      "loss": 16.9676,
+      "step": 7889
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.093124073433463e-05,
+      "loss": 17.5576,
+      "step": 7890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.092866003888479e-05,
+      "loss": 16.4421,
+      "step": 7891
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.092607901292507e-05,
+      "loss": 17.6321,
+      "step": 7892
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.092349765647633e-05,
+      "loss": 15.4525,
+      "step": 7893
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.092091596955941e-05,
+      "loss": 14.6483,
+      "step": 7894
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.091833395219517e-05,
+      "loss": 16.4778,
+      "step": 7895
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.091575160440447e-05,
+      "loss": 16.486,
+      "step": 7896
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.091316892620813e-05,
+      "loss": 13.2522,
+      "step": 7897
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.091058591762702e-05,
+      "loss": 16.6983,
+      "step": 7898
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.090800257868202e-05,
+      "loss": 16.1331,
+      "step": 7899
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.090541890939395e-05,
+      "loss": 15.224,
+      "step": 7900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.09028349097837e-05,
+      "loss": 15.6959,
+      "step": 7901
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.090025057987213e-05,
+      "loss": 18.351,
+      "step": 7902
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.089766591968012e-05,
+      "loss": 16.9335,
+      "step": 7903
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.089508092922853e-05,
+      "loss": 17.2369,
+      "step": 7904
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.089249560853825e-05,
+      "loss": 17.3822,
+      "step": 7905
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.088990995763012e-05,
+      "loss": 15.6402,
+      "step": 7906
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.088732397652506e-05,
+      "loss": 15.5976,
+      "step": 7907
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.088473766524394e-05,
+      "loss": 15.2575,
+      "step": 7908
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.088215102380764e-05,
+      "loss": 14.3438,
+      "step": 7909
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.087956405223704e-05,
+      "loss": 17.5795,
+      "step": 7910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.087697675055304e-05,
+      "loss": 15.7224,
+      "step": 7911
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.087438911877655e-05,
+      "loss": 19.3686,
+      "step": 7912
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.087180115692844e-05,
+      "loss": 17.1005,
+      "step": 7913
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.086921286502961e-05,
+      "loss": 14.9952,
+      "step": 7914
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.086662424310098e-05,
+      "loss": 16.9753,
+      "step": 7915
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.086403529116345e-05,
+      "loss": 15.5517,
+      "step": 7916
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.086144600923791e-05,
+      "loss": 17.0441,
+      "step": 7917
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.085885639734527e-05,
+      "loss": 16.3797,
+      "step": 7918
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.085626645550646e-05,
+      "loss": 14.7163,
+      "step": 7919
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.085367618374239e-05,
+      "loss": 15.6596,
+      "step": 7920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.085108558207395e-05,
+      "loss": 15.284,
+      "step": 7921
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.08484946505221e-05,
+      "loss": 16.5337,
+      "step": 7922
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.084590338910774e-05,
+      "loss": 16.1817,
+      "step": 7923
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.08433117978518e-05,
+      "loss": 17.0291,
+      "step": 7924
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.084071987677522e-05,
+      "loss": 16.5173,
+      "step": 7925
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.083812762589889e-05,
+      "loss": 14.5851,
+      "step": 7926
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.083553504524381e-05,
+      "loss": 15.2431,
+      "step": 7927
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.083294213483084e-05,
+      "loss": 15.9412,
+      "step": 7928
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.083034889468096e-05,
+      "loss": 15.7464,
+      "step": 7929
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.082775532481512e-05,
+      "loss": 17.1485,
+      "step": 7930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.082516142525425e-05,
+      "loss": 16.1686,
+      "step": 7931
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.08225671960193e-05,
+      "loss": 16.991,
+      "step": 7932
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.08199726371312e-05,
+      "loss": 16.8651,
+      "step": 7933
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.081737774861092e-05,
+      "loss": 15.6457,
+      "step": 7934
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.081478253047941e-05,
+      "loss": 16.5421,
+      "step": 7935
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.081218698275763e-05,
+      "loss": 17.2355,
+      "step": 7936
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.080959110546654e-05,
+      "loss": 15.2899,
+      "step": 7937
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.080699489862709e-05,
+      "loss": 15.1232,
+      "step": 7938
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.080439836226027e-05,
+      "loss": 15.4761,
+      "step": 7939
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.080180149638704e-05,
+      "loss": 16.3499,
+      "step": 7940
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.079920430102835e-05,
+      "loss": 15.4008,
+      "step": 7941
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.079660677620521e-05,
+      "loss": 15.3854,
+      "step": 7942
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.079400892193854e-05,
+      "loss": 17.8225,
+      "step": 7943
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.079141073824937e-05,
+      "loss": 17.5682,
+      "step": 7944
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.078881222515864e-05,
+      "loss": 19.56,
+      "step": 7945
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.07862133826874e-05,
+      "loss": 14.9014,
+      "step": 7946
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.078361421085655e-05,
+      "loss": 15.3591,
+      "step": 7947
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.078101470968715e-05,
+      "loss": 15.8651,
+      "step": 7948
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.077841487920014e-05,
+      "loss": 17.2216,
+      "step": 7949
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.077581471941655e-05,
+      "loss": 18.3642,
+      "step": 7950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.077321423035736e-05,
+      "loss": 16.1215,
+      "step": 7951
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.077061341204359e-05,
+      "loss": 16.9185,
+      "step": 7952
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.07680122644962e-05,
+      "loss": 17.9201,
+      "step": 7953
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.076541078773623e-05,
+      "loss": 16.6519,
+      "step": 7954
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.076280898178469e-05,
+      "loss": 15.6274,
+      "step": 7955
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.076020684666255e-05,
+      "loss": 17.6078,
+      "step": 7956
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.075760438239088e-05,
+      "loss": 14.519,
+      "step": 7957
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.075500158899067e-05,
+      "loss": 16.5679,
+      "step": 7958
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.075239846648291e-05,
+      "loss": 16.8069,
+      "step": 7959
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.074979501488867e-05,
+      "loss": 17.9338,
+      "step": 7960
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.074719123422894e-05,
+      "loss": 15.5315,
+      "step": 7961
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.074458712452476e-05,
+      "loss": 14.3872,
+      "step": 7962
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.074198268579715e-05,
+      "loss": 16.7061,
+      "step": 7963
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.073937791806715e-05,
+      "loss": 15.0884,
+      "step": 7964
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.073677282135579e-05,
+      "loss": 15.2496,
+      "step": 7965
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.07341673956841e-05,
+      "loss": 15.5303,
+      "step": 7966
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.073156164107314e-05,
+      "loss": 15.9947,
+      "step": 7967
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.072895555754393e-05,
+      "loss": 17.2839,
+      "step": 7968
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.072634914511753e-05,
+      "loss": 16.9382,
+      "step": 7969
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.072374240381498e-05,
+      "loss": 17.0326,
+      "step": 7970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.072113533365734e-05,
+      "loss": 16.33,
+      "step": 7971
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.071852793466564e-05,
+      "loss": 16.2008,
+      "step": 7972
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.071592020686097e-05,
+      "loss": 15.8461,
+      "step": 7973
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.071331215026435e-05,
+      "loss": 14.8803,
+      "step": 7974
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.071070376489688e-05,
+      "loss": 14.88,
+      "step": 7975
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.070809505077957e-05,
+      "loss": 16.1632,
+      "step": 7976
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.070548600793354e-05,
+      "loss": 16.5522,
+      "step": 7977
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.070287663637985e-05,
+      "loss": 16.0082,
+      "step": 7978
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.070026693613953e-05,
+      "loss": 14.8198,
+      "step": 7979
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.06976569072337e-05,
+      "loss": 15.007,
+      "step": 7980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.069504654968342e-05,
+      "loss": 16.4924,
+      "step": 7981
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.069243586350975e-05,
+      "loss": 15.6662,
+      "step": 7982
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.06898248487338e-05,
+      "loss": 16.0394,
+      "step": 7983
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.068721350537666e-05,
+      "loss": 15.6517,
+      "step": 7984
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.06846018334594e-05,
+      "loss": 16.4104,
+      "step": 7985
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.068198983300309e-05,
+      "loss": 14.2794,
+      "step": 7986
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.067937750402887e-05,
+      "loss": 18.1883,
+      "step": 7987
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.067676484655779e-05,
+      "loss": 14.9104,
+      "step": 7988
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.067415186061098e-05,
+      "loss": 17.2276,
+      "step": 7989
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.067153854620953e-05,
+      "loss": 15.9057,
+      "step": 7990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.066892490337453e-05,
+      "loss": 17.5921,
+      "step": 7991
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.066631093212712e-05,
+      "loss": 16.1419,
+      "step": 7992
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.066369663248837e-05,
+      "loss": 17.7375,
+      "step": 7993
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.06610820044794e-05,
+      "loss": 14.5926,
+      "step": 7994
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.065846704812135e-05,
+      "loss": 15.9592,
+      "step": 7995
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.065585176343531e-05,
+      "loss": 18.2579,
+      "step": 7996
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.065323615044241e-05,
+      "loss": 15.7843,
+      "step": 7997
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.065062020916377e-05,
+      "loss": 16.8904,
+      "step": 7998
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.06480039396205e-05,
+      "loss": 16.6166,
+      "step": 7999
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.064538734183375e-05,
+      "loss": 18.3615,
+      "step": 8000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.064277041582465e-05,
+      "loss": 17.0888,
+      "step": 8001
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.06401531616143e-05,
+      "loss": 15.089,
+      "step": 8002
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.063753557922388e-05,
+      "loss": 17.9933,
+      "step": 8003
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.063491766867449e-05,
+      "loss": 16.0934,
+      "step": 8004
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.063229942998728e-05,
+      "loss": 15.6377,
+      "step": 8005
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.062968086318342e-05,
+      "loss": 15.0144,
+      "step": 8006
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.0627061968284e-05,
+      "loss": 16.1108,
+      "step": 8007
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.062444274531024e-05,
+      "loss": 14.8116,
+      "step": 8008
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.062182319428325e-05,
+      "loss": 17.2033,
+      "step": 8009
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.061920331522416e-05,
+      "loss": 15.8941,
+      "step": 8010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.061658310815417e-05,
+      "loss": 15.0433,
+      "step": 8011
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.061396257309443e-05,
+      "loss": 14.9716,
+      "step": 8012
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.061134171006608e-05,
+      "loss": 15.8317,
+      "step": 8013
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.06087205190903e-05,
+      "loss": 15.8632,
+      "step": 8014
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.060609900018825e-05,
+      "loss": 14.4186,
+      "step": 8015
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.060347715338111e-05,
+      "loss": 16.2467,
+      "step": 8016
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.060085497869004e-05,
+      "loss": 15.7332,
+      "step": 8017
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.059823247613623e-05,
+      "loss": 15.8004,
+      "step": 8018
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.059560964574085e-05,
+      "loss": 17.1179,
+      "step": 8019
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.059298648752507e-05,
+      "loss": 17.2429,
+      "step": 8020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.059036300151007e-05,
+      "loss": 16.9696,
+      "step": 8021
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.058773918771706e-05,
+      "loss": 17.9862,
+      "step": 8022
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.058511504616723e-05,
+      "loss": 16.5784,
+      "step": 8023
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.058249057688172e-05,
+      "loss": 18.8159,
+      "step": 8024
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.057986577988179e-05,
+      "loss": 15.781,
+      "step": 8025
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.057724065518859e-05,
+      "loss": 16.6892,
+      "step": 8026
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.057461520282333e-05,
+      "loss": 16.1322,
+      "step": 8027
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.057198942280722e-05,
+      "loss": 16.4214,
+      "step": 8028
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.056936331516146e-05,
+      "loss": 15.6748,
+      "step": 8029
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.056673687990724e-05,
+      "loss": 15.2378,
+      "step": 8030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.05641101170658e-05,
+      "loss": 14.5223,
+      "step": 8031
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.056148302665832e-05,
+      "loss": 17.5192,
+      "step": 8032
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.055885560870605e-05,
+      "loss": 15.2,
+      "step": 8033
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.055622786323017e-05,
+      "loss": 16.0513,
+      "step": 8034
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.055359979025191e-05,
+      "loss": 15.9621,
+      "step": 8035
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.055097138979252e-05,
+      "loss": 16.0866,
+      "step": 8036
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.05483426618732e-05,
+      "loss": 16.4945,
+      "step": 8037
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.054571360651517e-05,
+      "loss": 14.4085,
+      "step": 8038
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.054308422373968e-05,
+      "loss": 18.1964,
+      "step": 8039
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.054045451356796e-05,
+      "loss": 18.1591,
+      "step": 8040
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.053782447602123e-05,
+      "loss": 15.4317,
+      "step": 8041
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.053519411112075e-05,
+      "loss": 16.5513,
+      "step": 8042
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.053256341888774e-05,
+      "loss": 16.6952,
+      "step": 8043
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.052993239934346e-05,
+      "loss": 17.4641,
+      "step": 8044
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.052730105250914e-05,
+      "loss": 16.0634,
+      "step": 8045
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.052466937840605e-05,
+      "loss": 14.6728,
+      "step": 8046
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.052203737705541e-05,
+      "loss": 15.9353,
+      "step": 8047
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.051940504847851e-05,
+      "loss": 15.4457,
+      "step": 8048
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.05167723926966e-05,
+      "loss": 15.9411,
+      "step": 8049
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.05141394097309e-05,
+      "loss": 15.6892,
+      "step": 8050
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.051150609960272e-05,
+      "loss": 15.6951,
+      "step": 8051
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.050887246233332e-05,
+      "loss": 17.6538,
+      "step": 8052
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.050623849794395e-05,
+      "loss": 16.6951,
+      "step": 8053
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.050360420645587e-05,
+      "loss": 15.8138,
+      "step": 8054
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.050096958789035e-05,
+      "loss": 16.7311,
+      "step": 8055
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.049833464226873e-05,
+      "loss": 14.6722,
+      "step": 8056
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.049569936961221e-05,
+      "loss": 15.014,
+      "step": 8057
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.049306376994212e-05,
+      "loss": 15.9035,
+      "step": 8058
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.04904278432797e-05,
+      "loss": 16.7595,
+      "step": 8059
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.048779158964627e-05,
+      "loss": 14.3306,
+      "step": 8060
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.048515500906309e-05,
+      "loss": 15.273,
+      "step": 8061
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.048251810155148e-05,
+      "loss": 16.9961,
+      "step": 8062
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.047988086713273e-05,
+      "loss": 14.814,
+      "step": 8063
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.047724330582813e-05,
+      "loss": 17.065,
+      "step": 8064
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.047460541765897e-05,
+      "loss": 16.2119,
+      "step": 8065
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.047196720264657e-05,
+      "loss": 15.5011,
+      "step": 8066
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.046932866081222e-05,
+      "loss": 14.8587,
+      "step": 8067
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.046668979217723e-05,
+      "loss": 17.2188,
+      "step": 8068
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.046405059676291e-05,
+      "loss": 17.227,
+      "step": 8069
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.046141107459056e-05,
+      "loss": 15.8827,
+      "step": 8070
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.045877122568152e-05,
+      "loss": 13.7198,
+      "step": 8071
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.045613105005711e-05,
+      "loss": 16.8844,
+      "step": 8072
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.04534905477386e-05,
+      "loss": 15.994,
+      "step": 8073
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.045084971874738e-05,
+      "loss": 15.1926,
+      "step": 8074
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.044820856310472e-05,
+      "loss": 18.4399,
+      "step": 8075
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.044556708083198e-05,
+      "loss": 16.4438,
+      "step": 8076
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.044292527195049e-05,
+      "loss": 15.7449,
+      "step": 8077
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.044028313648157e-05,
+      "loss": 15.8358,
+      "step": 8078
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.043764067444656e-05,
+      "loss": 15.6758,
+      "step": 8079
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.04349978858668e-05,
+      "loss": 15.9428,
+      "step": 8080
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.043235477076363e-05,
+      "loss": 15.5847,
+      "step": 8081
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.04297113291584e-05,
+      "loss": 13.9478,
+      "step": 8082
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.042706756107244e-05,
+      "loss": 17.1089,
+      "step": 8083
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.042442346652713e-05,
+      "loss": 17.3486,
+      "step": 8084
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.042177904554378e-05,
+      "loss": 16.6651,
+      "step": 8085
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.041913429814379e-05,
+      "loss": 15.8162,
+      "step": 8086
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.041648922434847e-05,
+      "loss": 15.7221,
+      "step": 8087
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.041384382417922e-05,
+      "loss": 13.5384,
+      "step": 8088
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.041119809765738e-05,
+      "loss": 17.7241,
+      "step": 8089
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.040855204480431e-05,
+      "loss": 15.4768,
+      "step": 8090
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.040590566564141e-05,
+      "loss": 16.7552,
+      "step": 8091
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.040325896019002e-05,
+      "loss": 17.1194,
+      "step": 8092
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.04006119284715e-05,
+      "loss": 14.7179,
+      "step": 8093
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.039796457050728e-05,
+      "loss": 16.054,
+      "step": 8094
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.03953168863187e-05,
+      "loss": 14.9365,
+      "step": 8095
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.039266887592714e-05,
+      "loss": 14.7471,
+      "step": 8096
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.039002053935398e-05,
+      "loss": 15.4771,
+      "step": 8097
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.038737187662062e-05,
+      "loss": 15.7715,
+      "step": 8098
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.038472288774843e-05,
+      "loss": 14.6167,
+      "step": 8099
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.038207357275885e-05,
+      "loss": 13.9036,
+      "step": 8100
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.037942393167322e-05,
+      "loss": 17.3289,
+      "step": 8101
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.037677396451295e-05,
+      "loss": 14.2218,
+      "step": 8102
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.037412367129945e-05,
+      "loss": 18.3958,
+      "step": 8103
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.037147305205412e-05,
+      "loss": 15.3952,
+      "step": 8104
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.036882210679835e-05,
+      "loss": 18.1925,
+      "step": 8105
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.036617083555357e-05,
+      "loss": 14.4481,
+      "step": 8106
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.036351923834117e-05,
+      "loss": 17.1029,
+      "step": 8107
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.036086731518259e-05,
+      "loss": 15.4956,
+      "step": 8108
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.03582150660992e-05,
+      "loss": 17.1106,
+      "step": 8109
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.035556249111246e-05,
+      "loss": 15.7661,
+      "step": 8110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.035290959024375e-05,
+      "loss": 16.3706,
+      "step": 8111
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.035025636351452e-05,
+      "loss": 15.7818,
+      "step": 8112
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.034760281094621e-05,
+      "loss": 15.3546,
+      "step": 8113
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.034494893256023e-05,
+      "loss": 17.4214,
+      "step": 8114
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.034229472837801e-05,
+      "loss": 17.372,
+      "step": 8115
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.033964019842096e-05,
+      "loss": 17.1951,
+      "step": 8116
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.033698534271055e-05,
+      "loss": 18.3782,
+      "step": 8117
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.033433016126822e-05,
+      "loss": 16.0741,
+      "step": 8118
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.033167465411541e-05,
+      "loss": 17.1531,
+      "step": 8119
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.032901882127354e-05,
+      "loss": 17.0974,
+      "step": 8120
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.032636266276407e-05,
+      "loss": 15.6278,
+      "step": 8121
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.032370617860844e-05,
+      "loss": 15.2702,
+      "step": 8122
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.032104936882811e-05,
+      "loss": 16.5908,
+      "step": 8123
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.031839223344456e-05,
+      "loss": 17.7079,
+      "step": 8124
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.03157347724792e-05,
+      "loss": 16.7481,
+      "step": 8125
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.031307698595353e-05,
+      "loss": 15.8236,
+      "step": 8126
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.031041887388898e-05,
+      "loss": 18.291,
+      "step": 8127
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.030776043630704e-05,
+      "loss": 17.2441,
+      "step": 8128
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.030510167322916e-05,
+      "loss": 17.0477,
+      "step": 8129
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.030244258467681e-05,
+      "loss": 16.7459,
+      "step": 8130
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.029978317067149e-05,
+      "loss": 16.0614,
+      "step": 8131
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.029712343123464e-05,
+      "loss": 15.7034,
+      "step": 8132
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.029446336638777e-05,
+      "loss": 17.1586,
+      "step": 8133
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.029180297615235e-05,
+      "loss": 16.0401,
+      "step": 8134
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.028914226054983e-05,
+      "loss": 15.8478,
+      "step": 8135
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.028648121960173e-05,
+      "loss": 15.7845,
+      "step": 8136
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.028381985332954e-05,
+      "loss": 15.8193,
+      "step": 8137
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.028115816175476e-05,
+      "loss": 17.4005,
+      "step": 8138
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.027849614489885e-05,
+      "loss": 17.3973,
+      "step": 8139
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.027583380278333e-05,
+      "loss": 16.2965,
+      "step": 8140
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.027317113542969e-05,
+      "loss": 18.0243,
+      "step": 8141
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.027050814285943e-05,
+      "loss": 16.3241,
+      "step": 8142
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.026784482509407e-05,
+      "loss": 15.8229,
+      "step": 8143
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.02651811821551e-05,
+      "loss": 13.401,
+      "step": 8144
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.026251721406404e-05,
+      "loss": 16.6037,
+      "step": 8145
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.02598529208424e-05,
+      "loss": 16.5638,
+      "step": 8146
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.025718830251169e-05,
+      "loss": 16.0497,
+      "step": 8147
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.025452335909343e-05,
+      "loss": 16.5924,
+      "step": 8148
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.025185809060915e-05,
+      "loss": 15.5989,
+      "step": 8149
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.024919249708035e-05,
+      "loss": 16.5343,
+      "step": 8150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.024652657852859e-05,
+      "loss": 14.0925,
+      "step": 8151
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.024386033497536e-05,
+      "loss": 15.1986,
+      "step": 8152
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.02411937664422e-05,
+      "loss": 17.931,
+      "step": 8153
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.023852687295066e-05,
+      "loss": 15.5466,
+      "step": 8154
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.023585965452228e-05,
+      "loss": 17.0055,
+      "step": 8155
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.023319211117858e-05,
+      "loss": 16.434,
+      "step": 8156
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.02305242429411e-05,
+      "loss": 18.6899,
+      "step": 8157
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.022785604983139e-05,
+      "loss": 16.0708,
+      "step": 8158
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.0225187531871e-05,
+      "loss": 15.7546,
+      "step": 8159
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.022251868908146e-05,
+      "loss": 16.2685,
+      "step": 8160
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.021984952148437e-05,
+      "loss": 18.5773,
+      "step": 8161
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.021718002910124e-05,
+      "loss": 15.462,
+      "step": 8162
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.021451021195362e-05,
+      "loss": 18.3795,
+      "step": 8163
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.021184007006311e-05,
+      "loss": 16.677,
+      "step": 8164
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.020916960345123e-05,
+      "loss": 18.1481,
+      "step": 8165
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.020649881213958e-05,
+      "loss": 16.1292,
+      "step": 8166
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.02038276961497e-05,
+      "loss": 15.6098,
+      "step": 8167
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.020115625550318e-05,
+      "loss": 15.6843,
+      "step": 8168
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.019848449022159e-05,
+      "loss": 14.6252,
+      "step": 8169
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.019581240032648e-05,
+      "loss": 17.0763,
+      "step": 8170
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.019313998583947e-05,
+      "loss": 17.0696,
+      "step": 8171
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.019046724678211e-05,
+      "loss": 16.6216,
+      "step": 8172
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.0187794183176e-05,
+      "loss": 15.263,
+      "step": 8173
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.018512079504272e-05,
+      "loss": 18.7267,
+      "step": 8174
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.018244708240382e-05,
+      "loss": 18.1582,
+      "step": 8175
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.017977304528095e-05,
+      "loss": 15.9134,
+      "step": 8176
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.017709868369568e-05,
+      "loss": 18.1063,
+      "step": 8177
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.017442399766961e-05,
+      "loss": 15.5789,
+      "step": 8178
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.017174898722433e-05,
+      "loss": 15.8091,
+      "step": 8179
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.016907365238146e-05,
+      "loss": 16.6081,
+      "step": 8180
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.016639799316256e-05,
+      "loss": 16.1014,
+      "step": 8181
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.016372200958929e-05,
+      "loss": 18.0904,
+      "step": 8182
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.016104570168322e-05,
+      "loss": 15.1419,
+      "step": 8183
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.0158369069466e-05,
+      "loss": 16.4687,
+      "step": 8184
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.015569211295921e-05,
+      "loss": 14.7614,
+      "step": 8185
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.015301483218447e-05,
+      "loss": 17.1518,
+      "step": 8186
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.015033722716344e-05,
+      "loss": 17.7352,
+      "step": 8187
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.014765929791768e-05,
+      "loss": 16.2313,
+      "step": 8188
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.014498104446887e-05,
+      "loss": 16.4566,
+      "step": 8189
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.01423024668386e-05,
+      "loss": 15.8864,
+      "step": 8190
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.013962356504851e-05,
+      "loss": 15.9893,
+      "step": 8191
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.013694433912025e-05,
+      "loss": 15.0971,
+      "step": 8192
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.013426478907545e-05,
+      "loss": 17.6571,
+      "step": 8193
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.013158491493572e-05,
+      "loss": 16.4084,
+      "step": 8194
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.012890471672272e-05,
+      "loss": 16.8309,
+      "step": 8195
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.012622419445812e-05,
+      "loss": 16.831,
+      "step": 8196
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.012354334816353e-05,
+      "loss": 17.7142,
+      "step": 8197
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.01208621778606e-05,
+      "loss": 17.4718,
+      "step": 8198
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.0118180683571e-05,
+      "loss": 17.1668,
+      "step": 8199
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.011549886531638e-05,
+      "loss": 15.8526,
+      "step": 8200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.011281672311838e-05,
+      "loss": 15.0026,
+      "step": 8201
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.011013425699868e-05,
+      "loss": 16.1086,
+      "step": 8202
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.010745146697892e-05,
+      "loss": 17.2238,
+      "step": 8203
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.010476835308079e-05,
+      "loss": 13.6022,
+      "step": 8204
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.010208491532594e-05,
+      "loss": 15.4053,
+      "step": 8205
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.009940115373604e-05,
+      "loss": 16.4763,
+      "step": 8206
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.009671706833276e-05,
+      "loss": 14.449,
+      "step": 8207
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.009403265913777e-05,
+      "loss": 13.7753,
+      "step": 8208
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.009134792617278e-05,
+      "loss": 15.268,
+      "step": 8209
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.008866286945943e-05,
+      "loss": 14.6561,
+      "step": 8210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.008597748901942e-05,
+      "loss": 17.2262,
+      "step": 8211
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.008329178487442e-05,
+      "loss": 14.6619,
+      "step": 8212
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.008060575704616e-05,
+      "loss": 18.5459,
+      "step": 8213
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.007791940555628e-05,
+      "loss": 16.7893,
+      "step": 8214
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.00752327304265e-05,
+      "loss": 16.6917,
+      "step": 8215
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.007254573167853e-05,
+      "loss": 15.4886,
+      "step": 8216
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.006985840933405e-05,
+      "loss": 15.9816,
+      "step": 8217
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.006717076341474e-05,
+      "loss": 16.2481,
+      "step": 8218
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.006448279394231e-05,
+      "loss": 17.1122,
+      "step": 8219
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.00617945009385e-05,
+      "loss": 16.2071,
+      "step": 8220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.005910588442499e-05,
+      "loss": 16.3612,
+      "step": 8221
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.005641694442352e-05,
+      "loss": 16.9923,
+      "step": 8222
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.005372768095576e-05,
+      "loss": 16.972,
+      "step": 8223
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.005103809404347e-05,
+      "loss": 17.3115,
+      "step": 8224
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.004834818370834e-05,
+      "loss": 17.5526,
+      "step": 8225
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.004565794997209e-05,
+      "loss": 17.2345,
+      "step": 8226
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.004296739285647e-05,
+      "loss": 18.3267,
+      "step": 8227
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.004027651238318e-05,
+      "loss": 16.5252,
+      "step": 8228
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.003758530857397e-05,
+      "loss": 17.8097,
+      "step": 8229
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.003489378145056e-05,
+      "loss": 17.2857,
+      "step": 8230
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.003220193103469e-05,
+      "loss": 16.0749,
+      "step": 8231
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.00295097573481e-05,
+      "loss": 17.3071,
+      "step": 8232
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.002681726041251e-05,
+      "loss": 15.1369,
+      "step": 8233
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.00241244402497e-05,
+      "loss": 16.6051,
+      "step": 8234
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.002143129688137e-05,
+      "loss": 15.2703,
+      "step": 8235
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.001873783032931e-05,
+      "loss": 15.2008,
+      "step": 8236
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.001604404061526e-05,
+      "loss": 17.0331,
+      "step": 8237
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.001334992776094e-05,
+      "loss": 16.3798,
+      "step": 8238
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.001065549178814e-05,
+      "loss": 16.0578,
+      "step": 8239
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.000796073271862e-05,
+      "loss": 17.5963,
+      "step": 8240
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.000526565057412e-05,
+      "loss": 16.8158,
+      "step": 8241
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.000257024537641e-05,
+      "loss": 16.3657,
+      "step": 8242
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.999987451714727e-05,
+      "loss": 14.806,
+      "step": 8243
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.999717846590846e-05,
+      "loss": 15.3389,
+      "step": 8244
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.999448209168175e-05,
+      "loss": 17.5574,
+      "step": 8245
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.99917853944889e-05,
+      "loss": 14.6291,
+      "step": 8246
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.998908837435171e-05,
+      "loss": 15.3959,
+      "step": 8247
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.998639103129196e-05,
+      "loss": 16.3603,
+      "step": 8248
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.998369336533141e-05,
+      "loss": 15.7814,
+      "step": 8249
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.998099537649184e-05,
+      "loss": 15.3383,
+      "step": 8250
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.997829706479508e-05,
+      "loss": 17.6623,
+      "step": 8251
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.997559843026288e-05,
+      "loss": 17.2429,
+      "step": 8252
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.997289947291705e-05,
+      "loss": 15.8599,
+      "step": 8253
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.997020019277938e-05,
+      "loss": 15.8114,
+      "step": 8254
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.996750058987164e-05,
+      "loss": 16.5049,
+      "step": 8255
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.996480066421568e-05,
+      "loss": 14.7148,
+      "step": 8256
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.996210041583329e-05,
+      "loss": 16.6984,
+      "step": 8257
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.995939984474624e-05,
+      "loss": 17.9913,
+      "step": 8258
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.995669895097638e-05,
+      "loss": 14.9316,
+      "step": 8259
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.995399773454549e-05,
+      "loss": 16.1506,
+      "step": 8260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.99512961954754e-05,
+      "loss": 17.2139,
+      "step": 8261
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.994859433378793e-05,
+      "loss": 16.8885,
+      "step": 8262
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.994589214950489e-05,
+      "loss": 14.1731,
+      "step": 8263
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.994318964264809e-05,
+      "loss": 15.3483,
+      "step": 8264
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.994048681323936e-05,
+      "loss": 15.3631,
+      "step": 8265
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.993778366130054e-05,
+      "loss": 16.5933,
+      "step": 8266
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.993508018685344e-05,
+      "loss": 16.9131,
+      "step": 8267
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.993237638991992e-05,
+      "loss": 16.2905,
+      "step": 8268
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.992967227052178e-05,
+      "loss": 17.9051,
+      "step": 8269
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.992696782868086e-05,
+      "loss": 16.5728,
+      "step": 8270
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.992426306441903e-05,
+      "loss": 15.2813,
+      "step": 8271
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.99215579777581e-05,
+      "loss": 17.4934,
+      "step": 8272
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.991885256871993e-05,
+      "loss": 15.0249,
+      "step": 8273
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.991614683732635e-05,
+      "loss": 15.9866,
+      "step": 8274
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.991344078359924e-05,
+      "loss": 15.4101,
+      "step": 8275
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.991073440756041e-05,
+      "loss": 15.2394,
+      "step": 8276
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.990802770923177e-05,
+      "loss": 16.8004,
+      "step": 8277
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.990532068863513e-05,
+      "loss": 16.7039,
+      "step": 8278
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.990261334579235e-05,
+      "loss": 18.0097,
+      "step": 8279
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.989990568072532e-05,
+      "loss": 14.7124,
+      "step": 8280
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.989719769345591e-05,
+      "loss": 17.0482,
+      "step": 8281
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.989448938400595e-05,
+      "loss": 17.1605,
+      "step": 8282
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.989178075239734e-05,
+      "loss": 17.0115,
+      "step": 8283
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.988907179865196e-05,
+      "loss": 17.0276,
+      "step": 8284
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.988636252279165e-05,
+      "loss": 16.5864,
+      "step": 8285
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.988365292483831e-05,
+      "loss": 16.5095,
+      "step": 8286
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.988094300481382e-05,
+      "loss": 14.9568,
+      "step": 8287
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.987823276274008e-05,
+      "loss": 16.419,
+      "step": 8288
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.987552219863893e-05,
+      "loss": 14.9973,
+      "step": 8289
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.98728113125323e-05,
+      "loss": 15.362,
+      "step": 8290
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.987010010444207e-05,
+      "loss": 16.0873,
+      "step": 8291
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.986738857439014e-05,
+      "loss": 16.3263,
+      "step": 8292
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.986467672239838e-05,
+      "loss": 16.4221,
+      "step": 8293
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.986196454848871e-05,
+      "loss": 16.4446,
+      "step": 8294
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.985925205268304e-05,
+      "loss": 16.0977,
+      "step": 8295
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.985653923500326e-05,
+      "loss": 17.9926,
+      "step": 8296
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.985382609547127e-05,
+      "loss": 15.8621,
+      "step": 8297
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.9851112634109e-05,
+      "loss": 15.6249,
+      "step": 8298
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.984839885093832e-05,
+      "loss": 15.4598,
+      "step": 8299
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.984568474598121e-05,
+      "loss": 14.944,
+      "step": 8300
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.984297031925954e-05,
+      "loss": 16.2311,
+      "step": 8301
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.984025557079523e-05,
+      "loss": 16.99,
+      "step": 8302
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.983754050061023e-05,
+      "loss": 16.2242,
+      "step": 8303
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.983482510872645e-05,
+      "loss": 15.8153,
+      "step": 8304
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.98321093951658e-05,
+      "loss": 17.4395,
+      "step": 8305
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.982939335995024e-05,
+      "loss": 15.5715,
+      "step": 8306
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.98266770031017e-05,
+      "loss": 15.5957,
+      "step": 8307
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.982396032464208e-05,
+      "loss": 15.75,
+      "step": 8308
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.982124332459336e-05,
+      "loss": 16.3084,
+      "step": 8309
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.981852600297745e-05,
+      "loss": 16.8323,
+      "step": 8310
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.98158083598163e-05,
+      "loss": 16.3202,
+      "step": 8311
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.98130903951319e-05,
+      "loss": 14.2029,
+      "step": 8312
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.981037210894614e-05,
+      "loss": 18.4291,
+      "step": 8313
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.980765350128098e-05,
+      "loss": 15.4067,
+      "step": 8314
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.98049345721584e-05,
+      "loss": 15.8414,
+      "step": 8315
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.980221532160033e-05,
+      "loss": 16.9977,
+      "step": 8316
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.979949574962874e-05,
+      "loss": 16.2862,
+      "step": 8317
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.979677585626559e-05,
+      "loss": 17.2866,
+      "step": 8318
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.979405564153285e-05,
+      "loss": 17.2539,
+      "step": 8319
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.979133510545248e-05,
+      "loss": 16.0744,
+      "step": 8320
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.978861424804644e-05,
+      "loss": 17.8775,
+      "step": 8321
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.978589306933671e-05,
+      "loss": 16.304,
+      "step": 8322
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.978317156934528e-05,
+      "loss": 15.9551,
+      "step": 8323
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.978044974809412e-05,
+      "loss": 17.5164,
+      "step": 8324
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.977772760560518e-05,
+      "loss": 15.393,
+      "step": 8325
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.977500514190047e-05,
+      "loss": 16.6081,
+      "step": 8326
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.977228235700196e-05,
+      "loss": 16.4,
+      "step": 8327
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.976955925093167e-05,
+      "loss": 17.0269,
+      "step": 8328
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.976683582371156e-05,
+      "loss": 17.5972,
+      "step": 8329
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.976411207536361e-05,
+      "loss": 15.049,
+      "step": 8330
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.976138800590986e-05,
+      "loss": 16.6519,
+      "step": 8331
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.975866361537227e-05,
+      "loss": 15.9396,
+      "step": 8332
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.975593890377284e-05,
+      "loss": 16.0221,
+      "step": 8333
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.97532138711336e-05,
+      "loss": 14.9971,
+      "step": 8334
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.975048851747653e-05,
+      "loss": 17.3807,
+      "step": 8335
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.974776284282365e-05,
+      "loss": 19.6359,
+      "step": 8336
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.974503684719697e-05,
+      "loss": 17.284,
+      "step": 8337
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.97423105306185e-05,
+      "loss": 16.5366,
+      "step": 8338
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.973958389311023e-05,
+      "loss": 15.2878,
+      "step": 8339
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.973685693469423e-05,
+      "loss": 16.6708,
+      "step": 8340
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.973412965539249e-05,
+      "loss": 17.2185,
+      "step": 8341
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.973140205522705e-05,
+      "loss": 18.2777,
+      "step": 8342
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.972867413421993e-05,
+      "loss": 17.1206,
+      "step": 8343
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.972594589239314e-05,
+      "loss": 14.7577,
+      "step": 8344
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.972321732976872e-05,
+      "loss": 16.5855,
+      "step": 8345
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.972048844636871e-05,
+      "loss": 15.5013,
+      "step": 8346
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.971775924221515e-05,
+      "loss": 14.6322,
+      "step": 8347
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.971502971733008e-05,
+      "loss": 15.4049,
+      "step": 8348
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.971229987173553e-05,
+      "loss": 17.3892,
+      "step": 8349
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.970956970545355e-05,
+      "loss": 16.3659,
+      "step": 8350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.97068392185062e-05,
+      "loss": 14.1311,
+      "step": 8351
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.97041084109155e-05,
+      "loss": 17.3817,
+      "step": 8352
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.970137728270353e-05,
+      "loss": 15.0949,
+      "step": 8353
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.969864583389234e-05,
+      "loss": 15.4611,
+      "step": 8354
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.969591406450397e-05,
+      "loss": 17.6769,
+      "step": 8355
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.96931819745605e-05,
+      "loss": 15.5172,
+      "step": 8356
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.969044956408396e-05,
+      "loss": 15.9806,
+      "step": 8357
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.968771683309645e-05,
+      "loss": 15.3516,
+      "step": 8358
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.968498378162003e-05,
+      "loss": 17.3728,
+      "step": 8359
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.968225040967676e-05,
+      "loss": 16.6345,
+      "step": 8360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.967951671728871e-05,
+      "loss": 14.4223,
+      "step": 8361
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.967678270447799e-05,
+      "loss": 16.9072,
+      "step": 8362
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.967404837126662e-05,
+      "loss": 14.8076,
+      "step": 8363
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.967131371767674e-05,
+      "loss": 17.1083,
+      "step": 8364
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.966857874373038e-05,
+      "loss": 16.4266,
+      "step": 8365
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.966584344944966e-05,
+      "loss": 15.297,
+      "step": 8366
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.966310783485667e-05,
+      "loss": 15.4804,
+      "step": 8367
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.966037189997347e-05,
+      "loss": 16.7329,
+      "step": 8368
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.965763564482219e-05,
+      "loss": 18.3486,
+      "step": 8369
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.96548990694249e-05,
+      "loss": 15.1856,
+      "step": 8370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.96521621738037e-05,
+      "loss": 19.2505,
+      "step": 8371
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.96494249579807e-05,
+      "loss": 14.67,
+      "step": 8372
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.964668742197801e-05,
+      "loss": 17.4976,
+      "step": 8373
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.964394956581774e-05,
+      "loss": 15.8943,
+      "step": 8374
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.964121138952197e-05,
+      "loss": 14.7554,
+      "step": 8375
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.963847289311285e-05,
+      "loss": 15.3868,
+      "step": 8376
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.963573407661245e-05,
+      "loss": 16.5136,
+      "step": 8377
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.963299494004291e-05,
+      "loss": 17.6978,
+      "step": 8378
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.963025548342636e-05,
+      "loss": 16.9055,
+      "step": 8379
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.962751570678492e-05,
+      "loss": 14.1321,
+      "step": 8380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.962477561014068e-05,
+      "loss": 15.6798,
+      "step": 8381
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.96220351935158e-05,
+      "loss": 16.8908,
+      "step": 8382
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.961929445693243e-05,
+      "loss": 15.1318,
+      "step": 8383
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.961655340041264e-05,
+      "loss": 16.4098,
+      "step": 8384
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.96138120239786e-05,
+      "loss": 18.4073,
+      "step": 8385
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.961107032765245e-05,
+      "loss": 17.1792,
+      "step": 8386
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.960832831145635e-05,
+      "loss": 16.6404,
+      "step": 8387
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.960558597541239e-05,
+      "loss": 17.1788,
+      "step": 8388
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.960284331954277e-05,
+      "loss": 16.9375,
+      "step": 8389
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.960010034386959e-05,
+      "loss": 17.1146,
+      "step": 8390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.959735704841503e-05,
+      "loss": 16.2874,
+      "step": 8391
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.959461343320123e-05,
+      "loss": 17.3364,
+      "step": 8392
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.959186949825035e-05,
+      "loss": 16.242,
+      "step": 8393
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.958912524358456e-05,
+      "loss": 16.6649,
+      "step": 8394
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.9586380669226e-05,
+      "loss": 17.0593,
+      "step": 8395
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.958363577519684e-05,
+      "loss": 15.9859,
+      "step": 8396
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.958089056151924e-05,
+      "loss": 15.1776,
+      "step": 8397
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.95781450282154e-05,
+      "loss": 18.0204,
+      "step": 8398
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.957539917530744e-05,
+      "loss": 17.8439,
+      "step": 8399
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.957265300281758e-05,
+      "loss": 17.4211,
+      "step": 8400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.956990651076796e-05,
+      "loss": 16.273,
+      "step": 8401
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.956715969918079e-05,
+      "loss": 15.6388,
+      "step": 8402
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.956441256807822e-05,
+      "loss": 15.2666,
+      "step": 8403
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.956166511748247e-05,
+      "loss": 16.1425,
+      "step": 8404
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.95589173474157e-05,
+      "loss": 16.1281,
+      "step": 8405
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.955616925790009e-05,
+      "loss": 16.5565,
+      "step": 8406
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.955342084895786e-05,
+      "loss": 17.6661,
+      "step": 8407
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.95506721206112e-05,
+      "loss": 14.9738,
+      "step": 8408
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.954792307288229e-05,
+      "loss": 15.6598,
+      "step": 8409
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.954517370579333e-05,
+      "loss": 17.5206,
+      "step": 8410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.954242401936652e-05,
+      "loss": 14.2975,
+      "step": 8411
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.95396740136241e-05,
+      "loss": 16.3415,
+      "step": 8412
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.953692368858824e-05,
+      "loss": 15.5939,
+      "step": 8413
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.953417304428116e-05,
+      "loss": 16.85,
+      "step": 8414
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.953142208072507e-05,
+      "loss": 15.0875,
+      "step": 8415
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.952867079794218e-05,
+      "loss": 14.0031,
+      "step": 8416
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.952591919595472e-05,
+      "loss": 16.782,
+      "step": 8417
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.95231672747849e-05,
+      "loss": 15.2255,
+      "step": 8418
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.952041503445496e-05,
+      "loss": 18.1132,
+      "step": 8419
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.95176624749871e-05,
+      "loss": 15.49,
+      "step": 8420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.951490959640357e-05,
+      "loss": 17.0588,
+      "step": 8421
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.951215639872659e-05,
+      "loss": 17.1948,
+      "step": 8422
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.950940288197838e-05,
+      "loss": 14.9007,
+      "step": 8423
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.95066490461812e-05,
+      "loss": 15.6513,
+      "step": 8424
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.950389489135727e-05,
+      "loss": 16.5172,
+      "step": 8425
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.950114041752884e-05,
+      "loss": 15.2968,
+      "step": 8426
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.949838562471814e-05,
+      "loss": 16.5274,
+      "step": 8427
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.949563051294743e-05,
+      "loss": 18.0948,
+      "step": 8428
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.949287508223896e-05,
+      "loss": 14.715,
+      "step": 8429
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.949011933261497e-05,
+      "loss": 15.0826,
+      "step": 8430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.948736326409771e-05,
+      "loss": 17.1115,
+      "step": 8431
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.948460687670945e-05,
+      "loss": 14.6797,
+      "step": 8432
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.948185017047245e-05,
+      "loss": 15.1971,
+      "step": 8433
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.947909314540895e-05,
+      "loss": 14.9734,
+      "step": 8434
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.947633580154125e-05,
+      "loss": 16.2834,
+      "step": 8435
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.947357813889157e-05,
+      "loss": 15.1145,
+      "step": 8436
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.947082015748221e-05,
+      "loss": 16.6752,
+      "step": 8437
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.946806185733543e-05,
+      "loss": 16.3334,
+      "step": 8438
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.946530323847351e-05,
+      "loss": 15.0408,
+      "step": 8439
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.946254430091873e-05,
+      "loss": 17.6756,
+      "step": 8440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.945978504469337e-05,
+      "loss": 15.9216,
+      "step": 8441
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.945702546981969e-05,
+      "loss": 16.0872,
+      "step": 8442
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.945426557632e-05,
+      "loss": 16.196,
+      "step": 8443
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.945150536421656e-05,
+      "loss": 14.9477,
+      "step": 8444
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.94487448335317e-05,
+      "loss": 15.4392,
+      "step": 8445
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.944598398428767e-05,
+      "loss": 16.474,
+      "step": 8446
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.944322281650677e-05,
+      "loss": 16.2253,
+      "step": 8447
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.944046133021133e-05,
+      "loss": 15.6215,
+      "step": 8448
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.943769952542363e-05,
+      "loss": 17.7421,
+      "step": 8449
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.943493740216598e-05,
+      "loss": 16.7202,
+      "step": 8450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.943217496046066e-05,
+      "loss": 16.4028,
+      "step": 8451
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.942941220033e-05,
+      "loss": 17.7214,
+      "step": 8452
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.94266491217963e-05,
+      "loss": 17.6359,
+      "step": 8453
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.942388572488187e-05,
+      "loss": 16.659,
+      "step": 8454
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.942112200960905e-05,
+      "loss": 16.4215,
+      "step": 8455
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.941835797600014e-05,
+      "loss": 16.1767,
+      "step": 8456
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.941559362407745e-05,
+      "loss": 15.7684,
+      "step": 8457
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.94128289538633e-05,
+      "loss": 15.6258,
+      "step": 8458
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.941006396538006e-05,
+      "loss": 15.7076,
+      "step": 8459
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.940729865865e-05,
+      "loss": 17.1739,
+      "step": 8460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.940453303369548e-05,
+      "loss": 16.479,
+      "step": 8461
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.940176709053884e-05,
+      "loss": 16.2591,
+      "step": 8462
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.939900082920239e-05,
+      "loss": 15.5449,
+      "step": 8463
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.939623424970851e-05,
+      "loss": 16.5716,
+      "step": 8464
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.93934673520795e-05,
+      "loss": 16.6643,
+      "step": 8465
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.939070013633771e-05,
+      "loss": 15.4063,
+      "step": 8466
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.93879326025055e-05,
+      "loss": 15.9256,
+      "step": 8467
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.938516475060522e-05,
+      "loss": 15.1128,
+      "step": 8468
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.938239658065921e-05,
+      "loss": 17.9246,
+      "step": 8469
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.937962809268982e-05,
+      "loss": 16.567,
+      "step": 8470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.937685928671942e-05,
+      "loss": 17.2265,
+      "step": 8471
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.937409016277037e-05,
+      "loss": 15.081,
+      "step": 8472
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.937132072086502e-05,
+      "loss": 16.0496,
+      "step": 8473
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.936855096102575e-05,
+      "loss": 18.9626,
+      "step": 8474
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.93657808832749e-05,
+      "loss": 16.4013,
+      "step": 8475
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.936301048763484e-05,
+      "loss": 16.345,
+      "step": 8476
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.936023977412797e-05,
+      "loss": 16.1583,
+      "step": 8477
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.935746874277667e-05,
+      "loss": 14.9832,
+      "step": 8478
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.935469739360327e-05,
+      "loss": 14.8907,
+      "step": 8479
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.935192572663018e-05,
+      "loss": 14.1438,
+      "step": 8480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.934915374187977e-05,
+      "loss": 14.7077,
+      "step": 8481
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.934638143937447e-05,
+      "loss": 17.4719,
+      "step": 8482
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.934360881913659e-05,
+      "loss": 15.0654,
+      "step": 8483
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.934083588118857e-05,
+      "loss": 14.6578,
+      "step": 8484
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.933806262555279e-05,
+      "loss": 16.2208,
+      "step": 8485
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.933528905225165e-05,
+      "loss": 16.444,
+      "step": 8486
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.933251516130757e-05,
+      "loss": 17.1225,
+      "step": 8487
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.932974095274289e-05,
+      "loss": 15.5577,
+      "step": 8488
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.932696642658004e-05,
+      "loss": 15.9411,
+      "step": 8489
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.932419158284147e-05,
+      "loss": 15.0246,
+      "step": 8490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.932141642154953e-05,
+      "loss": 17.2537,
+      "step": 8491
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.931864094272663e-05,
+      "loss": 17.8591,
+      "step": 8492
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.931586514639522e-05,
+      "loss": 16.2792,
+      "step": 8493
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.931308903257769e-05,
+      "loss": 16.2218,
+      "step": 8494
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.931031260129646e-05,
+      "loss": 15.2326,
+      "step": 8495
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.930753585257396e-05,
+      "loss": 14.6464,
+      "step": 8496
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.930475878643262e-05,
+      "loss": 16.6349,
+      "step": 8497
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.930198140289484e-05,
+      "loss": 17.0294,
+      "step": 8498
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.929920370198307e-05,
+      "loss": 15.0127,
+      "step": 8499
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.929642568371973e-05,
+      "loss": 16.4568,
+      "step": 8500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.929364734812725e-05,
+      "loss": 15.2367,
+      "step": 8501
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.929086869522806e-05,
+      "loss": 14.2816,
+      "step": 8502
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.928808972504463e-05,
+      "loss": 17.573,
+      "step": 8503
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.928531043759938e-05,
+      "loss": 16.6132,
+      "step": 8504
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.928253083291475e-05,
+      "loss": 16.8007,
+      "step": 8505
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.927975091101319e-05,
+      "loss": 18.7269,
+      "step": 8506
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.927697067191716e-05,
+      "loss": 16.1972,
+      "step": 8507
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.927419011564908e-05,
+      "loss": 16.4674,
+      "step": 8508
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.927140924223146e-05,
+      "loss": 14.1268,
+      "step": 8509
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.926862805168668e-05,
+      "loss": 14.7184,
+      "step": 8510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.926584654403724e-05,
+      "loss": 16.7858,
+      "step": 8511
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.926306471930563e-05,
+      "loss": 15.718,
+      "step": 8512
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.926028257751427e-05,
+      "loss": 15.5998,
+      "step": 8513
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.925750011868563e-05,
+      "loss": 17.0388,
+      "step": 8514
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.925471734284221e-05,
+      "loss": 17.274,
+      "step": 8515
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.925193425000645e-05,
+      "loss": 16.9394,
+      "step": 8516
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.924915084020083e-05,
+      "loss": 15.1644,
+      "step": 8517
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.924636711344784e-05,
+      "loss": 15.4077,
+      "step": 8518
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.924358306976996e-05,
+      "loss": 16.8764,
+      "step": 8519
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.924079870918964e-05,
+      "loss": 17.931,
+      "step": 8520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.923801403172941e-05,
+      "loss": 16.2149,
+      "step": 8521
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.923522903741173e-05,
+      "loss": 14.8588,
+      "step": 8522
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.92324437262591e-05,
+      "loss": 16.0655,
+      "step": 8523
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.9229658098294e-05,
+      "loss": 16.2228,
+      "step": 8524
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.922687215353893e-05,
+      "loss": 17.2627,
+      "step": 8525
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.92240858920164e-05,
+      "loss": 17.7791,
+      "step": 8526
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.922129931374888e-05,
+      "loss": 17.3901,
+      "step": 8527
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.921851241875891e-05,
+      "loss": 19.0778,
+      "step": 8528
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.921572520706896e-05,
+      "loss": 16.4294,
+      "step": 8529
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.921293767870157e-05,
+      "loss": 16.9601,
+      "step": 8530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.921014983367923e-05,
+      "loss": 16.5,
+      "step": 8531
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.920736167202446e-05,
+      "loss": 15.5783,
+      "step": 8532
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.920457319375976e-05,
+      "loss": 15.5544,
+      "step": 8533
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.920178439890765e-05,
+      "loss": 15.4749,
+      "step": 8534
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.919899528749068e-05,
+      "loss": 16.0346,
+      "step": 8535
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.919620585953134e-05,
+      "loss": 14.408,
+      "step": 8536
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.919341611505219e-05,
+      "loss": 16.2265,
+      "step": 8537
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.91906260540757e-05,
+      "loss": 16.0675,
+      "step": 8538
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.918783567662446e-05,
+      "loss": 15.3542,
+      "step": 8539
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.918504498272098e-05,
+      "loss": 16.1664,
+      "step": 8540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.918225397238777e-05,
+      "loss": 17.7151,
+      "step": 8541
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.917946264564743e-05,
+      "loss": 15.9635,
+      "step": 8542
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.917667100252244e-05,
+      "loss": 17.365,
+      "step": 8543
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.917387904303536e-05,
+      "loss": 15.4764,
+      "step": 8544
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.917108676720875e-05,
+      "loss": 16.4685,
+      "step": 8545
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.916829417506516e-05,
+      "loss": 17.5667,
+      "step": 8546
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.916550126662713e-05,
+      "loss": 15.9453,
+      "step": 8547
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.916270804191722e-05,
+      "loss": 15.239,
+      "step": 8548
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.915991450095797e-05,
+      "loss": 16.6065,
+      "step": 8549
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.915712064377195e-05,
+      "loss": 16.2108,
+      "step": 8550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.915432647038173e-05,
+      "loss": 16.3101,
+      "step": 8551
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.915153198080985e-05,
+      "loss": 15.5215,
+      "step": 8552
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.91487371750789e-05,
+      "loss": 15.533,
+      "step": 8553
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.914594205321145e-05,
+      "loss": 15.5624,
+      "step": 8554
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.914314661523004e-05,
+      "loss": 15.1461,
+      "step": 8555
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.914035086115728e-05,
+      "loss": 15.303,
+      "step": 8556
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.913755479101574e-05,
+      "loss": 14.2651,
+      "step": 8557
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.913475840482797e-05,
+      "loss": 17.0828,
+      "step": 8558
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.913196170261657e-05,
+      "loss": 14.4811,
+      "step": 8559
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.912916468440415e-05,
+      "loss": 17.6975,
+      "step": 8560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.912636735021326e-05,
+      "loss": 14.4846,
+      "step": 8561
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.912356970006651e-05,
+      "loss": 16.0169,
+      "step": 8562
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.912077173398646e-05,
+      "loss": 17.9777,
+      "step": 8563
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.911797345199575e-05,
+      "loss": 15.1131,
+      "step": 8564
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.911517485411696e-05,
+      "loss": 15.797,
+      "step": 8565
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.911237594037267e-05,
+      "loss": 16.4087,
+      "step": 8566
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.910957671078551e-05,
+      "loss": 15.5871,
+      "step": 8567
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.910677716537806e-05,
+      "loss": 15.3614,
+      "step": 8568
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.910397730417292e-05,
+      "loss": 15.2125,
+      "step": 8569
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.910117712719274e-05,
+      "loss": 17.6467,
+      "step": 8570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.909837663446012e-05,
+      "loss": 18.5786,
+      "step": 8571
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.909557582599766e-05,
+      "loss": 14.6438,
+      "step": 8572
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.909277470182796e-05,
+      "loss": 15.6837,
+      "step": 8573
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.908997326197369e-05,
+      "loss": 15.6363,
+      "step": 8574
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.908717150645743e-05,
+      "loss": 16.9122,
+      "step": 8575
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.908436943530181e-05,
+      "loss": 15.5574,
+      "step": 8576
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.908156704852948e-05,
+      "loss": 16.5345,
+      "step": 8577
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.907876434616304e-05,
+      "loss": 15.7542,
+      "step": 8578
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.907596132822516e-05,
+      "loss": 16.0187,
+      "step": 8579
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.907315799473846e-05,
+      "loss": 15.397,
+      "step": 8580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.907035434572554e-05,
+      "loss": 14.6852,
+      "step": 8581
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.90675503812091e-05,
+      "loss": 17.9966,
+      "step": 8582
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.906474610121175e-05,
+      "loss": 15.9645,
+      "step": 8583
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.906194150575614e-05,
+      "loss": 15.4461,
+      "step": 8584
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.905913659486491e-05,
+      "loss": 17.6824,
+      "step": 8585
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.905633136856072e-05,
+      "loss": 16.2,
+      "step": 8586
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.905352582686622e-05,
+      "loss": 17.1456,
+      "step": 8587
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.905071996980408e-05,
+      "loss": 15.2071,
+      "step": 8588
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.904791379739693e-05,
+      "loss": 15.5382,
+      "step": 8589
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.904510730966746e-05,
+      "loss": 14.2961,
+      "step": 8590
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.904230050663831e-05,
+      "loss": 16.0313,
+      "step": 8591
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.903949338833216e-05,
+      "loss": 16.6374,
+      "step": 8592
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.903668595477166e-05,
+      "loss": 17.6905,
+      "step": 8593
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.903387820597951e-05,
+      "loss": 16.5163,
+      "step": 8594
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.903107014197836e-05,
+      "loss": 16.8322,
+      "step": 8595
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.902826176279089e-05,
+      "loss": 15.9029,
+      "step": 8596
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.902545306843976e-05,
+      "loss": 16.3826,
+      "step": 8597
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.902264405894771e-05,
+      "loss": 17.1763,
+      "step": 8598
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.901983473433737e-05,
+      "loss": 15.7617,
+      "step": 8599
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.901702509463143e-05,
+      "loss": 14.2735,
+      "step": 8600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.901421513985261e-05,
+      "loss": 15.0977,
+      "step": 8601
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.901140487002358e-05,
+      "loss": 14.4408,
+      "step": 8602
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.900859428516702e-05,
+      "loss": 17.0821,
+      "step": 8603
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.900578338530565e-05,
+      "loss": 15.5281,
+      "step": 8604
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.900297217046215e-05,
+      "loss": 14.3385,
+      "step": 8605
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.900016064065923e-05,
+      "loss": 17.1773,
+      "step": 8606
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.899734879591961e-05,
+      "loss": 16.6817,
+      "step": 8607
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.899453663626598e-05,
+      "loss": 15.508,
+      "step": 8608
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.899172416172104e-05,
+      "loss": 14.6777,
+      "step": 8609
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.898891137230753e-05,
+      "loss": 17.9106,
+      "step": 8610
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.898609826804814e-05,
+      "loss": 16.3102,
+      "step": 8611
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.898328484896558e-05,
+      "loss": 16.4962,
+      "step": 8612
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.898047111508259e-05,
+      "loss": 15.4586,
+      "step": 8613
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.897765706642189e-05,
+      "loss": 17.3457,
+      "step": 8614
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.897484270300618e-05,
+      "loss": 17.5372,
+      "step": 8615
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.89720280248582e-05,
+      "loss": 17.4865,
+      "step": 8616
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.896921303200071e-05,
+      "loss": 16.6556,
+      "step": 8617
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.896639772445641e-05,
+      "loss": 17.7115,
+      "step": 8618
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.896358210224802e-05,
+      "loss": 15.9001,
+      "step": 8619
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.896076616539831e-05,
+      "loss": 17.4906,
+      "step": 8620
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.895794991393001e-05,
+      "loss": 15.4287,
+      "step": 8621
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.895513334786585e-05,
+      "loss": 16.4694,
+      "step": 8622
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.895231646722857e-05,
+      "loss": 16.2417,
+      "step": 8623
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.894949927204097e-05,
+      "loss": 15.9443,
+      "step": 8624
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.894668176232572e-05,
+      "loss": 16.5566,
+      "step": 8625
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.894386393810563e-05,
+      "loss": 14.9111,
+      "step": 8626
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.894104579940343e-05,
+      "loss": 15.4568,
+      "step": 8627
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.893822734624188e-05,
+      "loss": 16.1135,
+      "step": 8628
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.893540857864375e-05,
+      "loss": 15.8423,
+      "step": 8629
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.893258949663179e-05,
+      "loss": 15.548,
+      "step": 8630
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.892977010022879e-05,
+      "loss": 15.8954,
+      "step": 8631
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.892695038945746e-05,
+      "loss": 16.0587,
+      "step": 8632
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.892413036434064e-05,
+      "loss": 15.5705,
+      "step": 8633
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.892131002490105e-05,
+      "loss": 15.383,
+      "step": 8634
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.891848937116149e-05,
+      "loss": 17.8072,
+      "step": 8635
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.891566840314473e-05,
+      "loss": 14.02,
+      "step": 8636
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.891284712087358e-05,
+      "loss": 16.5233,
+      "step": 8637
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.891002552437076e-05,
+      "loss": 16.1225,
+      "step": 8638
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.890720361365909e-05,
+      "loss": 16.3996,
+      "step": 8639
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.890438138876138e-05,
+      "loss": 16.6112,
+      "step": 8640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.89015588497004e-05,
+      "loss": 16.3784,
+      "step": 8641
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.889873599649892e-05,
+      "loss": 15.9036,
+      "step": 8642
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.889591282917977e-05,
+      "loss": 17.3652,
+      "step": 8643
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.889308934776572e-05,
+      "loss": 17.579,
+      "step": 8644
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.88902655522796e-05,
+      "loss": 16.106,
+      "step": 8645
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.88874414427442e-05,
+      "loss": 16.9162,
+      "step": 8646
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.888461701918231e-05,
+      "loss": 16.7382,
+      "step": 8647
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.888179228161676e-05,
+      "loss": 17.0478,
+      "step": 8648
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.887896723007034e-05,
+      "loss": 16.3179,
+      "step": 8649
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.887614186456588e-05,
+      "loss": 17.8703,
+      "step": 8650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.88733161851262e-05,
+      "loss": 14.6621,
+      "step": 8651
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.887049019177409e-05,
+      "loss": 13.9212,
+      "step": 8652
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.88676638845324e-05,
+      "loss": 14.3462,
+      "step": 8653
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.886483726342394e-05,
+      "loss": 13.7868,
+      "step": 8654
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.886201032847153e-05,
+      "loss": 15.6199,
+      "step": 8655
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.885918307969802e-05,
+      "loss": 16.7757,
+      "step": 8656
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.885635551712621e-05,
+      "loss": 17.6537,
+      "step": 8657
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.885352764077897e-05,
+      "loss": 15.3396,
+      "step": 8658
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.88506994506791e-05,
+      "loss": 16.5812,
+      "step": 8659
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.884787094684946e-05,
+      "loss": 16.4735,
+      "step": 8660
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.884504212931288e-05,
+      "loss": 14.1782,
+      "step": 8661
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.884221299809221e-05,
+      "loss": 15.2872,
+      "step": 8662
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.883938355321029e-05,
+      "loss": 15.1234,
+      "step": 8663
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.883655379468998e-05,
+      "loss": 16.1564,
+      "step": 8664
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.883372372255414e-05,
+      "loss": 17.7414,
+      "step": 8665
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.883089333682558e-05,
+      "loss": 17.977,
+      "step": 8666
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.88280626375272e-05,
+      "loss": 15.1456,
+      "step": 8667
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.882523162468184e-05,
+      "loss": 15.6258,
+      "step": 8668
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.882240029831236e-05,
+      "loss": 15.856,
+      "step": 8669
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.881956865844163e-05,
+      "loss": 16.6301,
+      "step": 8670
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.88167367050925e-05,
+      "loss": 17.0648,
+      "step": 8671
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.881390443828787e-05,
+      "loss": 15.5987,
+      "step": 8672
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.88110718580506e-05,
+      "loss": 16.4587,
+      "step": 8673
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.880823896440353e-05,
+      "loss": 17.2358,
+      "step": 8674
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.880540575736959e-05,
+      "loss": 16.6016,
+      "step": 8675
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.880257223697163e-05,
+      "loss": 17.3685,
+      "step": 8676
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.879973840323252e-05,
+      "loss": 16.5435,
+      "step": 8677
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.879690425617517e-05,
+      "loss": 15.4221,
+      "step": 8678
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.879406979582246e-05,
+      "loss": 14.0031,
+      "step": 8679
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.879123502219724e-05,
+      "loss": 16.7057,
+      "step": 8680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.878839993532247e-05,
+      "loss": 14.6905,
+      "step": 8681
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.8785564535221e-05,
+      "loss": 17.6471,
+      "step": 8682
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.878272882191573e-05,
+      "loss": 17.5935,
+      "step": 8683
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.877989279542957e-05,
+      "loss": 15.0411,
+      "step": 8684
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.877705645578541e-05,
+      "loss": 18.6242,
+      "step": 8685
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.877421980300617e-05,
+      "loss": 15.2152,
+      "step": 8686
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.877138283711475e-05,
+      "loss": 14.8098,
+      "step": 8687
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.876854555813406e-05,
+      "loss": 15.1904,
+      "step": 8688
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.876570796608698e-05,
+      "loss": 17.0643,
+      "step": 8689
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.876287006099648e-05,
+      "loss": 16.7982,
+      "step": 8690
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.876003184288544e-05,
+      "loss": 15.5278,
+      "step": 8691
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.87571933117768e-05,
+      "loss": 15.1838,
+      "step": 8692
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.875435446769346e-05,
+      "loss": 14.9153,
+      "step": 8693
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.875151531065837e-05,
+      "loss": 15.5704,
+      "step": 8694
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.874867584069441e-05,
+      "loss": 15.9559,
+      "step": 8695
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.874583605782458e-05,
+      "loss": 17.0963,
+      "step": 8696
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.874299596207176e-05,
+      "loss": 15.8438,
+      "step": 8697
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.874015555345889e-05,
+      "loss": 14.8589,
+      "step": 8698
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.873731483200893e-05,
+      "loss": 16.1869,
+      "step": 8699
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.873447379774479e-05,
+      "loss": 14.6803,
+      "step": 8700
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.873163245068945e-05,
+      "loss": 16.3087,
+      "step": 8701
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.872879079086581e-05,
+      "loss": 14.9467,
+      "step": 8702
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.872594881829684e-05,
+      "loss": 18.2557,
+      "step": 8703
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.87231065330055e-05,
+      "loss": 16.4167,
+      "step": 8704
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.872026393501473e-05,
+      "loss": 14.551,
+      "step": 8705
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.871742102434748e-05,
+      "loss": 18.0298,
+      "step": 8706
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.871457780102671e-05,
+      "loss": 15.9456,
+      "step": 8707
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.871173426507539e-05,
+      "loss": 17.138,
+      "step": 8708
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.870889041651647e-05,
+      "loss": 15.3028,
+      "step": 8709
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.870604625537292e-05,
+      "loss": 16.9956,
+      "step": 8710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.870320178166772e-05,
+      "loss": 15.1914,
+      "step": 8711
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.870035699542381e-05,
+      "loss": 17.2061,
+      "step": 8712
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.869751189666418e-05,
+      "loss": 16.8441,
+      "step": 8713
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.869466648541182e-05,
+      "loss": 16.4474,
+      "step": 8714
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.869182076168969e-05,
+      "loss": 17.4008,
+      "step": 8715
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.868897472552076e-05,
+      "loss": 17.1299,
+      "step": 8716
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.868612837692802e-05,
+      "loss": 15.1155,
+      "step": 8717
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.868328171593448e-05,
+      "loss": 17.7606,
+      "step": 8718
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.868043474256309e-05,
+      "loss": 15.936,
+      "step": 8719
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.867758745683687e-05,
+      "loss": 16.0742,
+      "step": 8720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.867473985877879e-05,
+      "loss": 15.7385,
+      "step": 8721
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.867189194841185e-05,
+      "loss": 17.055,
+      "step": 8722
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.866904372575908e-05,
+      "loss": 16.2111,
+      "step": 8723
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.866619519084342e-05,
+      "loss": 15.9129,
+      "step": 8724
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.866334634368793e-05,
+      "loss": 16.021,
+      "step": 8725
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.866049718431558e-05,
+      "loss": 16.1045,
+      "step": 8726
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.865764771274938e-05,
+      "loss": 14.5243,
+      "step": 8727
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.865479792901236e-05,
+      "loss": 15.6288,
+      "step": 8728
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.865194783312751e-05,
+      "loss": 16.8071,
+      "step": 8729
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.864909742511787e-05,
+      "loss": 15.5171,
+      "step": 8730
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.864624670500643e-05,
+      "loss": 15.4773,
+      "step": 8731
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.864339567281623e-05,
+      "loss": 14.6324,
+      "step": 8732
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.86405443285703e-05,
+      "loss": 15.291,
+      "step": 8733
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.863769267229163e-05,
+      "loss": 14.443,
+      "step": 8734
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.86348407040033e-05,
+      "loss": 16.8284,
+      "step": 8735
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.863198842372828e-05,
+      "loss": 16.5167,
+      "step": 8736
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.862913583148966e-05,
+      "loss": 16.0044,
+      "step": 8737
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.862628292731043e-05,
+      "loss": 15.2001,
+      "step": 8738
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.862342971121367e-05,
+      "loss": 16.7003,
+      "step": 8739
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.862057618322238e-05,
+      "loss": 14.1208,
+      "step": 8740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.861772234335963e-05,
+      "loss": 15.6097,
+      "step": 8741
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.861486819164844e-05,
+      "loss": 14.0072,
+      "step": 8742
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.86120137281119e-05,
+      "loss": 15.282,
+      "step": 8743
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.860915895277301e-05,
+      "loss": 15.2628,
+      "step": 8744
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.860630386565487e-05,
+      "loss": 14.3763,
+      "step": 8745
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.860344846678051e-05,
+      "loss": 14.9866,
+      "step": 8746
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.860059275617297e-05,
+      "loss": 15.6664,
+      "step": 8747
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.859773673385535e-05,
+      "loss": 16.4534,
+      "step": 8748
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.85948803998507e-05,
+      "loss": 17.1788,
+      "step": 8749
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.859202375418207e-05,
+      "loss": 15.5746,
+      "step": 8750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.858916679687255e-05,
+      "loss": 17.1966,
+      "step": 8751
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.85863095279452e-05,
+      "loss": 17.1006,
+      "step": 8752
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.858345194742308e-05,
+      "loss": 15.2071,
+      "step": 8753
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.858059405532928e-05,
+      "loss": 15.7074,
+      "step": 8754
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.857773585168689e-05,
+      "loss": 16.9665,
+      "step": 8755
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.857487733651897e-05,
+      "loss": 15.2531,
+      "step": 8756
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.857201850984862e-05,
+      "loss": 15.5435,
+      "step": 8757
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.85691593716989e-05,
+      "loss": 15.3931,
+      "step": 8758
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.856629992209293e-05,
+      "loss": 16.019,
+      "step": 8759
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.856344016105378e-05,
+      "loss": 16.9625,
+      "step": 8760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.856058008860454e-05,
+      "loss": 15.597,
+      "step": 8761
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.855771970476834e-05,
+      "loss": 14.9164,
+      "step": 8762
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.855485900956825e-05,
+      "loss": 14.2556,
+      "step": 8763
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.855199800302736e-05,
+      "loss": 16.6605,
+      "step": 8764
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.85491366851688e-05,
+      "loss": 16.3747,
+      "step": 8765
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.854627505601566e-05,
+      "loss": 17.7303,
+      "step": 8766
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.854341311559105e-05,
+      "loss": 16.9827,
+      "step": 8767
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.854055086391808e-05,
+      "loss": 16.6486,
+      "step": 8768
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.853768830101987e-05,
+      "loss": 14.2264,
+      "step": 8769
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.853482542691953e-05,
+      "loss": 17.4002,
+      "step": 8770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.853196224164019e-05,
+      "loss": 14.989,
+      "step": 8771
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.852909874520496e-05,
+      "loss": 14.9256,
+      "step": 8772
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.852623493763697e-05,
+      "loss": 15.6014,
+      "step": 8773
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.852337081895933e-05,
+      "loss": 16.4415,
+      "step": 8774
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.852050638919518e-05,
+      "loss": 17.0061,
+      "step": 8775
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.851764164836765e-05,
+      "loss": 15.5554,
+      "step": 8776
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.851477659649989e-05,
+      "loss": 14.9021,
+      "step": 8777
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.851191123361502e-05,
+      "loss": 15.6041,
+      "step": 8778
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.850904555973616e-05,
+      "loss": 16.7323,
+      "step": 8779
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.850617957488648e-05,
+      "loss": 16.2082,
+      "step": 8780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.850331327908912e-05,
+      "loss": 15.3377,
+      "step": 8781
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.850044667236722e-05,
+      "loss": 15.3152,
+      "step": 8782
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.849757975474393e-05,
+      "loss": 19.728,
+      "step": 8783
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.849471252624239e-05,
+      "loss": 17.2998,
+      "step": 8784
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.849184498688578e-05,
+      "loss": 15.8108,
+      "step": 8785
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.848897713669721e-05,
+      "loss": 16.8862,
+      "step": 8786
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.84861089756999e-05,
+      "loss": 14.8291,
+      "step": 8787
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.848324050391697e-05,
+      "loss": 17.7728,
+      "step": 8788
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.848037172137159e-05,
+      "loss": 15.008,
+      "step": 8789
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.847750262808692e-05,
+      "loss": 15.8865,
+      "step": 8790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.847463322408616e-05,
+      "loss": 16.9038,
+      "step": 8791
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.847176350939244e-05,
+      "loss": 16.0276,
+      "step": 8792
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.846889348402896e-05,
+      "loss": 15.029,
+      "step": 8793
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.846602314801889e-05,
+      "loss": 18.2249,
+      "step": 8794
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.846315250138542e-05,
+      "loss": 17.9295,
+      "step": 8795
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.84602815441517e-05,
+      "loss": 14.3458,
+      "step": 8796
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.845741027634093e-05,
+      "loss": 15.3651,
+      "step": 8797
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.845453869797631e-05,
+      "loss": 17.4735,
+      "step": 8798
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.845166680908103e-05,
+      "loss": 15.7314,
+      "step": 8799
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.844879460967826e-05,
+      "loss": 15.6746,
+      "step": 8800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.844592209979119e-05,
+      "loss": 16.3933,
+      "step": 8801
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.844304927944304e-05,
+      "loss": 15.9556,
+      "step": 8802
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.8440176148657e-05,
+      "loss": 16.6222,
+      "step": 8803
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.843730270745628e-05,
+      "loss": 17.666,
+      "step": 8804
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.843442895586405e-05,
+      "loss": 14.9075,
+      "step": 8805
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.843155489390354e-05,
+      "loss": 16.3587,
+      "step": 8806
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.842868052159798e-05,
+      "loss": 14.7677,
+      "step": 8807
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.842580583897054e-05,
+      "loss": 15.4944,
+      "step": 8808
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.842293084604447e-05,
+      "loss": 18.0104,
+      "step": 8809
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.842005554284296e-05,
+      "loss": 15.5629,
+      "step": 8810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.841717992938924e-05,
+      "loss": 16.3921,
+      "step": 8811
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.841430400570652e-05,
+      "loss": 15.3609,
+      "step": 8812
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.841142777181805e-05,
+      "loss": 15.8387,
+      "step": 8813
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.840855122774703e-05,
+      "loss": 16.8601,
+      "step": 8814
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.840567437351671e-05,
+      "loss": 15.346,
+      "step": 8815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.84027972091503e-05,
+      "loss": 14.9164,
+      "step": 8816
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.839991973467105e-05,
+      "loss": 16.8726,
+      "step": 8817
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.839704195010219e-05,
+      "loss": 16.4364,
+      "step": 8818
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.839416385546696e-05,
+      "loss": 15.3239,
+      "step": 8819
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.839128545078859e-05,
+      "loss": 15.6683,
+      "step": 8820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.838840673609035e-05,
+      "loss": 15.2093,
+      "step": 8821
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.838552771139547e-05,
+      "loss": 14.9813,
+      "step": 8822
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.83826483767272e-05,
+      "loss": 18.3691,
+      "step": 8823
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.837976873210879e-05,
+      "loss": 14.9983,
+      "step": 8824
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.837688877756348e-05,
+      "loss": 17.5218,
+      "step": 8825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.837400851311458e-05,
+      "loss": 16.5676,
+      "step": 8826
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.837112793878527e-05,
+      "loss": 15.0223,
+      "step": 8827
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.836824705459887e-05,
+      "loss": 16.0081,
+      "step": 8828
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.836536586057862e-05,
+      "loss": 15.8615,
+      "step": 8829
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.83624843567478e-05,
+      "loss": 16.3106,
+      "step": 8830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.835960254312967e-05,
+      "loss": 15.3091,
+      "step": 8831
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.835672041974749e-05,
+      "loss": 15.7453,
+      "step": 8832
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.835383798662456e-05,
+      "loss": 16.1683,
+      "step": 8833
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.835095524378414e-05,
+      "loss": 15.8674,
+      "step": 8834
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.834807219124952e-05,
+      "loss": 14.9638,
+      "step": 8835
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.834518882904396e-05,
+      "loss": 16.222,
+      "step": 8836
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.834230515719078e-05,
+      "loss": 15.3451,
+      "step": 8837
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.833942117571321e-05,
+      "loss": 17.5063,
+      "step": 8838
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.833653688463459e-05,
+      "loss": 15.4667,
+      "step": 8839
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.833365228397821e-05,
+      "loss": 16.141,
+      "step": 8840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.833076737376733e-05,
+      "loss": 16.5444,
+      "step": 8841
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.832788215402527e-05,
+      "loss": 14.6979,
+      "step": 8842
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.832499662477532e-05,
+      "loss": 16.3222,
+      "step": 8843
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.832211078604078e-05,
+      "loss": 15.3229,
+      "step": 8844
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.831922463784497e-05,
+      "loss": 16.3346,
+      "step": 8845
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.831633818021117e-05,
+      "loss": 15.4585,
+      "step": 8846
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.831345141316272e-05,
+      "loss": 16.4552,
+      "step": 8847
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.831056433672291e-05,
+      "loss": 17.3771,
+      "step": 8848
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.830767695091506e-05,
+      "loss": 15.6151,
+      "step": 8849
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.830478925576248e-05,
+      "loss": 14.5735,
+      "step": 8850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.83019012512885e-05,
+      "loss": 15.7787,
+      "step": 8851
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.829901293751643e-05,
+      "loss": 14.2559,
+      "step": 8852
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.82961243144696e-05,
+      "loss": 16.3461,
+      "step": 8853
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.829323538217136e-05,
+      "loss": 15.5698,
+      "step": 8854
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.829034614064499e-05,
+      "loss": 14.3837,
+      "step": 8855
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.828745658991386e-05,
+      "loss": 17.3336,
+      "step": 8856
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.828456673000129e-05,
+      "loss": 16.3813,
+      "step": 8857
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.82816765609306e-05,
+      "loss": 17.1932,
+      "step": 8858
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.827878608272517e-05,
+      "loss": 16.0749,
+      "step": 8859
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.82758952954083e-05,
+      "loss": 17.5732,
+      "step": 8860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.827300419900337e-05,
+      "loss": 15.606,
+      "step": 8861
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.827011279353369e-05,
+      "loss": 16.2389,
+      "step": 8862
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.826722107902263e-05,
+      "loss": 15.2723,
+      "step": 8863
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.826432905549353e-05,
+      "loss": 15.349,
+      "step": 8864
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.826143672296976e-05,
+      "loss": 14.9458,
+      "step": 8865
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.825854408147467e-05,
+      "loss": 16.5595,
+      "step": 8866
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.825565113103162e-05,
+      "loss": 16.0083,
+      "step": 8867
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.825275787166397e-05,
+      "loss": 16.4398,
+      "step": 8868
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.824986430339506e-05,
+      "loss": 16.2675,
+      "step": 8869
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.824697042624829e-05,
+      "loss": 14.0866,
+      "step": 8870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.824407624024704e-05,
+      "loss": 14.7701,
+      "step": 8871
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.824118174541464e-05,
+      "loss": 17.7328,
+      "step": 8872
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.823828694177449e-05,
+      "loss": 17.0055,
+      "step": 8873
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.823539182934994e-05,
+      "loss": 17.1887,
+      "step": 8874
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.82324964081644e-05,
+      "loss": 14.9306,
+      "step": 8875
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.822960067824124e-05,
+      "loss": 17.1051,
+      "step": 8876
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.822670463960385e-05,
+      "loss": 15.6371,
+      "step": 8877
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.822380829227559e-05,
+      "loss": 18.1954,
+      "step": 8878
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.822091163627989e-05,
+      "loss": 14.9446,
+      "step": 8879
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.82180146716401e-05,
+      "loss": 15.5372,
+      "step": 8880
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.821511739837963e-05,
+      "loss": 17.434,
+      "step": 8881
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.821221981652189e-05,
+      "loss": 14.9912,
+      "step": 8882
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.820932192609026e-05,
+      "loss": 15.908,
+      "step": 8883
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.820642372710817e-05,
+      "loss": 15.8739,
+      "step": 8884
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.820352521959898e-05,
+      "loss": 15.4567,
+      "step": 8885
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.820062640358612e-05,
+      "loss": 16.434,
+      "step": 8886
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.819772727909301e-05,
+      "loss": 18.1851,
+      "step": 8887
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.819482784614304e-05,
+      "loss": 16.3411,
+      "step": 8888
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.819192810475964e-05,
+      "loss": 18.1689,
+      "step": 8889
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.81890280549662e-05,
+      "loss": 15.0135,
+      "step": 8890
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.818612769678618e-05,
+      "loss": 14.7797,
+      "step": 8891
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.818322703024297e-05,
+      "loss": 17.5037,
+      "step": 8892
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.818032605536e-05,
+      "loss": 16.719,
+      "step": 8893
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.817742477216072e-05,
+      "loss": 15.4895,
+      "step": 8894
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.81745231806685e-05,
+      "loss": 16.4443,
+      "step": 8895
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.817162128090684e-05,
+      "loss": 16.6748,
+      "step": 8896
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.816871907289913e-05,
+      "loss": 16.1714,
+      "step": 8897
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.81658165566688e-05,
+      "loss": 14.6394,
+      "step": 8898
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.816291373223933e-05,
+      "loss": 16.4421,
+      "step": 8899
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.816001059963412e-05,
+      "loss": 16.1084,
+      "step": 8900
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.815710715887664e-05,
+      "loss": 17.2919,
+      "step": 8901
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.815420340999033e-05,
+      "loss": 16.4273,
+      "step": 8902
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.815129935299864e-05,
+      "loss": 16.3642,
+      "step": 8903
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.8148394987925e-05,
+      "loss": 14.6262,
+      "step": 8904
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.81454903147929e-05,
+      "loss": 16.5572,
+      "step": 8905
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.814258533362576e-05,
+      "loss": 15.9253,
+      "step": 8906
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.813968004444706e-05,
+      "loss": 16.9506,
+      "step": 8907
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.813677444728026e-05,
+      "loss": 15.3802,
+      "step": 8908
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.813386854214882e-05,
+      "loss": 14.9964,
+      "step": 8909
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.81309623290762e-05,
+      "loss": 17.4895,
+      "step": 8910
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.812805580808588e-05,
+      "loss": 15.2457,
+      "step": 8911
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.81251489792013e-05,
+      "loss": 16.4684,
+      "step": 8912
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.812224184244599e-05,
+      "loss": 17.8283,
+      "step": 8913
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.811933439784339e-05,
+      "loss": 18.4153,
+      "step": 8914
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.811642664541697e-05,
+      "loss": 14.8778,
+      "step": 8915
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.811351858519023e-05,
+      "loss": 15.1933,
+      "step": 8916
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.811061021718664e-05,
+      "loss": 17.6352,
+      "step": 8917
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.810770154142969e-05,
+      "loss": 16.0414,
+      "step": 8918
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.810479255794288e-05,
+      "loss": 16.5993,
+      "step": 8919
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.810188326674969e-05,
+      "loss": 17.0317,
+      "step": 8920
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.80989736678736e-05,
+      "loss": 16.3828,
+      "step": 8921
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.809606376133815e-05,
+      "loss": 15.9488,
+      "step": 8922
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.809315354716679e-05,
+      "loss": 18.4617,
+      "step": 8923
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.809024302538304e-05,
+      "loss": 15.7611,
+      "step": 8924
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.80873321960104e-05,
+      "loss": 17.8981,
+      "step": 8925
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.80844210590724e-05,
+      "loss": 14.3332,
+      "step": 8926
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.808150961459251e-05,
+      "loss": 15.6677,
+      "step": 8927
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.807859786259425e-05,
+      "loss": 16.3556,
+      "step": 8928
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.807568580310114e-05,
+      "loss": 14.3408,
+      "step": 8929
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.807277343613672e-05,
+      "loss": 14.9101,
+      "step": 8930
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.806986076172447e-05,
+      "loss": 16.4064,
+      "step": 8931
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.806694777988792e-05,
+      "loss": 16.5294,
+      "step": 8932
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.80640344906506e-05,
+      "loss": 16.7997,
+      "step": 8933
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.806112089403604e-05,
+      "loss": 18.1036,
+      "step": 8934
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.805820699006776e-05,
+      "loss": 15.9521,
+      "step": 8935
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.80552927787693e-05,
+      "loss": 15.0671,
+      "step": 8936
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.805237826016417e-05,
+      "loss": 16.8004,
+      "step": 8937
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.804946343427594e-05,
+      "loss": 16.1934,
+      "step": 8938
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.804654830112811e-05,
+      "loss": 16.6385,
+      "step": 8939
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.804363286074423e-05,
+      "loss": 14.7252,
+      "step": 8940
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.804071711314788e-05,
+      "loss": 15.8687,
+      "step": 8941
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.803780105836257e-05,
+      "loss": 15.0926,
+      "step": 8942
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.803488469641185e-05,
+      "loss": 16.1452,
+      "step": 8943
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.803196802731927e-05,
+      "loss": 16.1609,
+      "step": 8944
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.802905105110842e-05,
+      "loss": 15.1723,
+      "step": 8945
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.802613376780278e-05,
+      "loss": 16.5876,
+      "step": 8946
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.802321617742598e-05,
+      "loss": 17.2477,
+      "step": 8947
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.802029828000156e-05,
+      "loss": 18.3849,
+      "step": 8948
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.801738007555304e-05,
+      "loss": 17.4107,
+      "step": 8949
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.801446156410405e-05,
+      "loss": 16.9541,
+      "step": 8950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.801154274567813e-05,
+      "loss": 15.8808,
+      "step": 8951
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.800862362029885e-05,
+      "loss": 15.184,
+      "step": 8952
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.800570418798976e-05,
+      "loss": 17.7316,
+      "step": 8953
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.800278444877448e-05,
+      "loss": 16.7279,
+      "step": 8954
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.799986440267655e-05,
+      "loss": 16.9484,
+      "step": 8955
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.799694404971957e-05,
+      "loss": 16.2071,
+      "step": 8956
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.79940233899271e-05,
+      "loss": 18.1244,
+      "step": 8957
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.799110242332276e-05,
+      "loss": 17.2362,
+      "step": 8958
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.798818114993011e-05,
+      "loss": 14.7753,
+      "step": 8959
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.798525956977277e-05,
+      "loss": 15.0697,
+      "step": 8960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.79823376828743e-05,
+      "loss": 16.6287,
+      "step": 8961
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.797941548925829e-05,
+      "loss": 15.7882,
+      "step": 8962
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.797649298894838e-05,
+      "loss": 17.5087,
+      "step": 8963
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.797357018196813e-05,
+      "loss": 14.1601,
+      "step": 8964
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.797064706834117e-05,
+      "loss": 16.8795,
+      "step": 8965
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.796772364809108e-05,
+      "loss": 17.1571,
+      "step": 8966
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.79647999212415e-05,
+      "loss": 17.0505,
+      "step": 8967
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.796187588781599e-05,
+      "loss": 16.0143,
+      "step": 8968
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.795895154783819e-05,
+      "loss": 16.5506,
+      "step": 8969
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.795602690133176e-05,
+      "loss": 16.2261,
+      "step": 8970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.795310194832023e-05,
+      "loss": 16.5284,
+      "step": 8971
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.795017668882728e-05,
+      "loss": 15.2437,
+      "step": 8972
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.794725112287651e-05,
+      "loss": 16.1575,
+      "step": 8973
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.794432525049156e-05,
+      "loss": 15.2486,
+      "step": 8974
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.794139907169605e-05,
+      "loss": 15.9651,
+      "step": 8975
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.79384725865136e-05,
+      "loss": 15.0775,
+      "step": 8976
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.793554579496786e-05,
+      "loss": 16.4685,
+      "step": 8977
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.793261869708244e-05,
+      "loss": 14.9191,
+      "step": 8978
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.7929691292881e-05,
+      "loss": 15.0773,
+      "step": 8979
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.792676358238718e-05,
+      "loss": 16.0537,
+      "step": 8980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.792383556562459e-05,
+      "loss": 16.2676,
+      "step": 8981
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.792090724261691e-05,
+      "loss": 17.8281,
+      "step": 8982
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.791797861338777e-05,
+      "loss": 16.5471,
+      "step": 8983
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.791504967796084e-05,
+      "loss": 15.9032,
+      "step": 8984
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.791212043635975e-05,
+      "loss": 15.4253,
+      "step": 8985
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.790919088860814e-05,
+      "loss": 15.7309,
+      "step": 8986
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.79062610347297e-05,
+      "loss": 15.1838,
+      "step": 8987
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.79033308747481e-05,
+      "loss": 16.2843,
+      "step": 8988
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.790040040868694e-05,
+      "loss": 16.4494,
+      "step": 8989
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.789746963656994e-05,
+      "loss": 14.8118,
+      "step": 8990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.789453855842076e-05,
+      "loss": 16.2352,
+      "step": 8991
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.789160717426303e-05,
+      "loss": 15.311,
+      "step": 8992
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.788867548412047e-05,
+      "loss": 16.958,
+      "step": 8993
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.788574348801675e-05,
+      "loss": 16.3673,
+      "step": 8994
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.78828111859755e-05,
+      "loss": 14.9826,
+      "step": 8995
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.787987857802045e-05,
+      "loss": 18.8279,
+      "step": 8996
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.787694566417526e-05,
+      "loss": 15.2172,
+      "step": 8997
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.787401244446361e-05,
+      "loss": 15.5632,
+      "step": 8998
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.78710789189092e-05,
+      "loss": 16.3918,
+      "step": 8999
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.78681450875357e-05,
+      "loss": 15.7709,
+      "step": 9000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.786521095036682e-05,
+      "loss": 16.0375,
+      "step": 9001
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.786227650742624e-05,
+      "loss": 16.8425,
+      "step": 9002
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.785934175873768e-05,
+      "loss": 16.5674,
+      "step": 9003
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.785640670432481e-05,
+      "loss": 14.5217,
+      "step": 9004
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.785347134421134e-05,
+      "loss": 16.2288,
+      "step": 9005
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.785053567842098e-05,
+      "loss": 16.4791,
+      "step": 9006
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.784759970697745e-05,
+      "loss": 15.6845,
+      "step": 9007
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.784466342990443e-05,
+      "loss": 14.0965,
+      "step": 9008
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.784172684722565e-05,
+      "loss": 14.8758,
+      "step": 9009
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.78387899589648e-05,
+      "loss": 17.7155,
+      "step": 9010
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.783585276514564e-05,
+      "loss": 16.3193,
+      "step": 9011
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.783291526579184e-05,
+      "loss": 15.4625,
+      "step": 9012
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.782997746092714e-05,
+      "loss": 16.571,
+      "step": 9013
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.782703935057529e-05,
+      "loss": 14.6654,
+      "step": 9014
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.782410093475999e-05,
+      "loss": 14.5227,
+      "step": 9015
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.782116221350494e-05,
+      "loss": 18.101,
+      "step": 9016
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.781822318683394e-05,
+      "loss": 18.1856,
+      "step": 9017
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.781528385477067e-05,
+      "loss": 15.8724,
+      "step": 9018
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.781234421733888e-05,
+      "loss": 16.1388,
+      "step": 9019
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.780940427456231e-05,
+      "loss": 16.7742,
+      "step": 9020
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.78064640264647e-05,
+      "loss": 19.5101,
+      "step": 9021
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.78035234730698e-05,
+      "loss": 17.2696,
+      "step": 9022
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.780058261440134e-05,
+      "loss": 15.9995,
+      "step": 9023
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.779764145048308e-05,
+      "loss": 14.0224,
+      "step": 9024
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.779469998133878e-05,
+      "loss": 15.5144,
+      "step": 9025
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.779175820699217e-05,
+      "loss": 17.5351,
+      "step": 9026
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.778881612746703e-05,
+      "loss": 16.8759,
+      "step": 9027
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.77858737427871e-05,
+      "loss": 16.6705,
+      "step": 9028
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.778293105297615e-05,
+      "loss": 15.8524,
+      "step": 9029
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.777998805805793e-05,
+      "loss": 16.8941,
+      "step": 9030
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.777704475805621e-05,
+      "loss": 17.3373,
+      "step": 9031
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.777410115299478e-05,
+      "loss": 15.9806,
+      "step": 9032
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.777115724289737e-05,
+      "loss": 15.4184,
+      "step": 9033
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.77682130277878e-05,
+      "loss": 18.153,
+      "step": 9034
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.776526850768981e-05,
+      "loss": 16.2897,
+      "step": 9035
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.776232368262719e-05,
+      "loss": 14.9715,
+      "step": 9036
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.775937855262372e-05,
+      "loss": 16.8258,
+      "step": 9037
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.775643311770318e-05,
+      "loss": 16.0272,
+      "step": 9038
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.775348737788936e-05,
+      "loss": 16.1278,
+      "step": 9039
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.775054133320604e-05,
+      "loss": 14.366,
+      "step": 9040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.774759498367702e-05,
+      "loss": 17.1882,
+      "step": 9041
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.774464832932609e-05,
+      "loss": 15.844,
+      "step": 9042
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.774170137017704e-05,
+      "loss": 14.8049,
+      "step": 9043
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.773875410625366e-05,
+      "loss": 15.8499,
+      "step": 9044
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.773580653757975e-05,
+      "loss": 18.8626,
+      "step": 9045
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.773285866417915e-05,
+      "loss": 18.5366,
+      "step": 9046
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.772991048607563e-05,
+      "loss": 16.4871,
+      "step": 9047
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.7726962003293e-05,
+      "loss": 15.6535,
+      "step": 9048
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.772401321585508e-05,
+      "loss": 15.1704,
+      "step": 9049
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.772106412378566e-05,
+      "loss": 15.8411,
+      "step": 9050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.771811472710858e-05,
+      "loss": 16.6885,
+      "step": 9051
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.771516502584765e-05,
+      "loss": 15.2066,
+      "step": 9052
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.77122150200267e-05,
+      "loss": 13.9532,
+      "step": 9053
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.770926470966953e-05,
+      "loss": 15.5303,
+      "step": 9054
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.770631409479996e-05,
+      "loss": 17.153,
+      "step": 9055
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.770336317544185e-05,
+      "loss": 18.7193,
+      "step": 9056
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.770041195161901e-05,
+      "loss": 17.4953,
+      "step": 9057
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.769746042335527e-05,
+      "loss": 16.313,
+      "step": 9058
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.769450859067448e-05,
+      "loss": 14.1328,
+      "step": 9059
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.769155645360043e-05,
+      "loss": 16.2557,
+      "step": 9060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.768860401215701e-05,
+      "loss": 17.0625,
+      "step": 9061
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.768565126636806e-05,
+      "loss": 17.358,
+      "step": 9062
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.768269821625741e-05,
+      "loss": 17.2292,
+      "step": 9063
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.767974486184889e-05,
+      "loss": 14.8887,
+      "step": 9064
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.767679120316636e-05,
+      "loss": 15.6946,
+      "step": 9065
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.76738372402337e-05,
+      "loss": 17.8074,
+      "step": 9066
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.767088297307472e-05,
+      "loss": 16.8645,
+      "step": 9067
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.76679284017133e-05,
+      "loss": 14.9021,
+      "step": 9068
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.766497352617331e-05,
+      "loss": 16.0289,
+      "step": 9069
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.766201834647858e-05,
+      "loss": 15.9607,
+      "step": 9070
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.7659062862653e-05,
+      "loss": 16.4802,
+      "step": 9071
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.765610707472043e-05,
+      "loss": 16.8195,
+      "step": 9072
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.765315098270474e-05,
+      "loss": 15.7648,
+      "step": 9073
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.76501945866298e-05,
+      "loss": 15.8873,
+      "step": 9074
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.764723788651947e-05,
+      "loss": 17.8227,
+      "step": 9075
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.764428088239764e-05,
+      "loss": 17.9618,
+      "step": 9076
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.764132357428818e-05,
+      "loss": 16.45,
+      "step": 9077
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.7638365962215e-05,
+      "loss": 16.069,
+      "step": 9078
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.763540804620194e-05,
+      "loss": 18.8547,
+      "step": 9079
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.763244982627291e-05,
+      "loss": 16.1768,
+      "step": 9080
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.762949130245179e-05,
+      "loss": 16.4356,
+      "step": 9081
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.762653247476249e-05,
+      "loss": 16.7477,
+      "step": 9082
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.762357334322889e-05,
+      "loss": 16.4378,
+      "step": 9083
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.762061390787488e-05,
+      "loss": 15.1689,
+      "step": 9084
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.761765416872436e-05,
+      "loss": 15.9261,
+      "step": 9085
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.761469412580125e-05,
+      "loss": 16.7712,
+      "step": 9086
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.761173377912942e-05,
+      "loss": 18.3929,
+      "step": 9087
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.760877312873279e-05,
+      "loss": 15.684,
+      "step": 9088
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.760581217463528e-05,
+      "loss": 14.2305,
+      "step": 9089
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.76028509168608e-05,
+      "loss": 17.3027,
+      "step": 9090
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.759988935543324e-05,
+      "loss": 15.8664,
+      "step": 9091
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.759692749037654e-05,
+      "loss": 17.4063,
+      "step": 9092
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.75939653217146e-05,
+      "loss": 20.335,
+      "step": 9093
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.759100284947135e-05,
+      "loss": 14.8818,
+      "step": 9094
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.75880400736707e-05,
+      "loss": 15.7762,
+      "step": 9095
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.75850769943366e-05,
+      "loss": 16.0419,
+      "step": 9096
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.758211361149295e-05,
+      "loss": 17.136,
+      "step": 9097
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.757914992516371e-05,
+      "loss": 16.1138,
+      "step": 9098
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.75761859353728e-05,
+      "loss": 15.8236,
+      "step": 9099
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.757322164214413e-05,
+      "loss": 14.1786,
+      "step": 9100
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.757025704550168e-05,
+      "loss": 16.2336,
+      "step": 9101
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.756729214546935e-05,
+      "loss": 15.7577,
+      "step": 9102
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.75643269420711e-05,
+      "loss": 15.8501,
+      "step": 9103
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.75613614353309e-05,
+      "loss": 15.3138,
+      "step": 9104
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.755839562527266e-05,
+      "loss": 16.7355,
+      "step": 9105
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.755542951192034e-05,
+      "loss": 16.1437,
+      "step": 9106
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.75524630952979e-05,
+      "loss": 17.7689,
+      "step": 9107
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.754949637542928e-05,
+      "loss": 15.8527,
+      "step": 9108
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.754652935233844e-05,
+      "loss": 18.3252,
+      "step": 9109
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.754356202604937e-05,
+      "loss": 16.6421,
+      "step": 9110
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.754059439658599e-05,
+      "loss": 17.2006,
+      "step": 9111
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.753762646397229e-05,
+      "loss": 14.0132,
+      "step": 9112
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.753465822823222e-05,
+      "loss": 15.1276,
+      "step": 9113
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.753168968938977e-05,
+      "loss": 17.2142,
+      "step": 9114
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.752872084746891e-05,
+      "loss": 15.4926,
+      "step": 9115
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.752575170249357e-05,
+      "loss": 15.1093,
+      "step": 9116
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.75227822544878e-05,
+      "loss": 15.9049,
+      "step": 9117
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.751981250347552e-05,
+      "loss": 18.125,
+      "step": 9118
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.751684244948073e-05,
+      "loss": 15.272,
+      "step": 9119
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.751387209252741e-05,
+      "loss": 17.5309,
+      "step": 9120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.751090143263957e-05,
+      "loss": 16.8001,
+      "step": 9121
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.750793046984116e-05,
+      "loss": 14.8215,
+      "step": 9122
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.750495920415623e-05,
+      "loss": 15.0054,
+      "step": 9123
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.75019876356087e-05,
+      "loss": 15.5077,
+      "step": 9124
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.749901576422262e-05,
+      "loss": 16.9808,
+      "step": 9125
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.749604359002196e-05,
+      "loss": 17.5508,
+      "step": 9126
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.749307111303073e-05,
+      "loss": 15.8899,
+      "step": 9127
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.749009833327295e-05,
+      "loss": 17.177,
+      "step": 9128
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.74871252507726e-05,
+      "loss": 16.2552,
+      "step": 9129
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.748415186555369e-05,
+      "loss": 14.7887,
+      "step": 9130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.748117817764028e-05,
+      "loss": 15.9659,
+      "step": 9131
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.74782041870563e-05,
+      "loss": 15.8919,
+      "step": 9132
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.747522989382584e-05,
+      "loss": 18.2293,
+      "step": 9133
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.747225529797287e-05,
+      "loss": 17.7557,
+      "step": 9134
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.746928039952143e-05,
+      "loss": 13.9592,
+      "step": 9135
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.746630519849555e-05,
+      "loss": 14.8278,
+      "step": 9136
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.746332969491925e-05,
+      "loss": 16.5185,
+      "step": 9137
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.746035388881655e-05,
+      "loss": 16.1436,
+      "step": 9138
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.745737778021149e-05,
+      "loss": 16.9477,
+      "step": 9139
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.74544013691281e-05,
+      "loss": 15.263,
+      "step": 9140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.74514246555904e-05,
+      "loss": 15.9889,
+      "step": 9141
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.744844763962245e-05,
+      "loss": 15.9005,
+      "step": 9142
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.74454703212483e-05,
+      "loss": 15.9134,
+      "step": 9143
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.744249270049196e-05,
+      "loss": 14.3106,
+      "step": 9144
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.743951477737749e-05,
+      "loss": 15.8724,
+      "step": 9145
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.743653655192891e-05,
+      "loss": 14.5769,
+      "step": 9146
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.743355802417033e-05,
+      "loss": 14.5473,
+      "step": 9147
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.743057919412577e-05,
+      "loss": 16.8289,
+      "step": 9148
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.742760006181929e-05,
+      "loss": 18.2551,
+      "step": 9149
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.742462062727492e-05,
+      "loss": 13.4988,
+      "step": 9150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.742164089051674e-05,
+      "loss": 17.1413,
+      "step": 9151
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.741866085156883e-05,
+      "loss": 16.6283,
+      "step": 9152
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.741568051045524e-05,
+      "loss": 17.2433,
+      "step": 9153
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.741269986720002e-05,
+      "loss": 17.4852,
+      "step": 9154
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.740971892182726e-05,
+      "loss": 16.7411,
+      "step": 9155
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.740673767436104e-05,
+      "loss": 14.4838,
+      "step": 9156
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.740375612482541e-05,
+      "loss": 14.4056,
+      "step": 9157
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.740077427324446e-05,
+      "loss": 17.5459,
+      "step": 9158
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.739779211964226e-05,
+      "loss": 15.5801,
+      "step": 9159
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.739480966404291e-05,
+      "loss": 15.0924,
+      "step": 9160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.739182690647048e-05,
+      "loss": 14.7884,
+      "step": 9161
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.738884384694905e-05,
+      "loss": 16.0243,
+      "step": 9162
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.738586048550272e-05,
+      "loss": 14.0653,
+      "step": 9163
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.73828768221556e-05,
+      "loss": 14.6785,
+      "step": 9164
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.737989285693176e-05,
+      "loss": 16.4804,
+      "step": 9165
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.737690858985527e-05,
+      "loss": 16.1827,
+      "step": 9166
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.737392402095028e-05,
+      "loss": 14.9938,
+      "step": 9167
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.737093915024088e-05,
+      "loss": 15.9569,
+      "step": 9168
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.736795397775114e-05,
+      "loss": 15.8536,
+      "step": 9169
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.736496850350522e-05,
+      "loss": 15.5365,
+      "step": 9170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.736198272752718e-05,
+      "loss": 18.6414,
+      "step": 9171
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.735899664984115e-05,
+      "loss": 14.7792,
+      "step": 9172
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.735601027047125e-05,
+      "loss": 14.2306,
+      "step": 9173
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.735302358944157e-05,
+      "loss": 16.3686,
+      "step": 9174
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.735003660677628e-05,
+      "loss": 14.4196,
+      "step": 9175
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.734704932249944e-05,
+      "loss": 17.0106,
+      "step": 9176
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.734406173663521e-05,
+      "loss": 18.2773,
+      "step": 9177
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.73410738492077e-05,
+      "loss": 15.6765,
+      "step": 9178
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.733808566024104e-05,
+      "loss": 16.2289,
+      "step": 9179
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.733509716975938e-05,
+      "loss": 17.3208,
+      "step": 9180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.733210837778683e-05,
+      "loss": 16.9413,
+      "step": 9181
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.732911928434753e-05,
+      "loss": 15.1026,
+      "step": 9182
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.732612988946562e-05,
+      "loss": 17.0571,
+      "step": 9183
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.732314019316524e-05,
+      "loss": 16.4546,
+      "step": 9184
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.732015019547055e-05,
+      "loss": 16.8799,
+      "step": 9185
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.731715989640565e-05,
+      "loss": 16.5425,
+      "step": 9186
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.731416929599474e-05,
+      "loss": 15.7224,
+      "step": 9187
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.731117839426193e-05,
+      "loss": 13.9997,
+      "step": 9188
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.73081871912314e-05,
+      "loss": 14.1346,
+      "step": 9189
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 8.730519568692728e-05,
+      "loss": 17.6221,
+      "step": 9190
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.730220388137375e-05,
+      "loss": 15.4774,
+      "step": 9191
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.729921177459496e-05,
+      "loss": 16.3373,
+      "step": 9192
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.729621936661507e-05,
+      "loss": 18.1093,
+      "step": 9193
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.729322665745824e-05,
+      "loss": 16.8014,
+      "step": 9194
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.729023364714865e-05,
+      "loss": 15.7863,
+      "step": 9195
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.728724033571045e-05,
+      "loss": 15.7324,
+      "step": 9196
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.728424672316784e-05,
+      "loss": 15.2189,
+      "step": 9197
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.728125280954498e-05,
+      "loss": 16.1398,
+      "step": 9198
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.727825859486602e-05,
+      "loss": 17.3013,
+      "step": 9199
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.727526407915518e-05,
+      "loss": 18.2036,
+      "step": 9200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.727226926243663e-05,
+      "loss": 14.6556,
+      "step": 9201
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.726927414473456e-05,
+      "loss": 16.7069,
+      "step": 9202
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.726627872607312e-05,
+      "loss": 15.8882,
+      "step": 9203
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.726328300647654e-05,
+      "loss": 16.4685,
+      "step": 9204
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.7260286985969e-05,
+      "loss": 17.7535,
+      "step": 9205
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.725729066457467e-05,
+      "loss": 16.453,
+      "step": 9206
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.725429404231777e-05,
+      "loss": 16.9073,
+      "step": 9207
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.72512971192225e-05,
+      "loss": 14.9328,
+      "step": 9208
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.724829989531307e-05,
+      "loss": 14.1996,
+      "step": 9209
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.724530237061366e-05,
+      "loss": 18.1784,
+      "step": 9210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.724230454514847e-05,
+      "loss": 15.4339,
+      "step": 9211
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.723930641894173e-05,
+      "loss": 16.0309,
+      "step": 9212
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.723630799201765e-05,
+      "loss": 16.2917,
+      "step": 9213
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.723330926440045e-05,
+      "loss": 15.2845,
+      "step": 9214
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.72303102361143e-05,
+      "loss": 18.5078,
+      "step": 9215
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.722731090718347e-05,
+      "loss": 15.7884,
+      "step": 9216
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.722431127763216e-05,
+      "loss": 15.8993,
+      "step": 9217
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.722131134748458e-05,
+      "loss": 18.3272,
+      "step": 9218
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.721831111676498e-05,
+      "loss": 15.651,
+      "step": 9219
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.721531058549757e-05,
+      "loss": 15.5121,
+      "step": 9220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.721230975370659e-05,
+      "loss": 14.6715,
+      "step": 9221
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.720930862141626e-05,
+      "loss": 16.8208,
+      "step": 9222
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.720630718865084e-05,
+      "loss": 16.5991,
+      "step": 9223
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.720330545543453e-05,
+      "loss": 17.3176,
+      "step": 9224
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.72003034217916e-05,
+      "loss": 14.8395,
+      "step": 9225
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.719730108774629e-05,
+      "loss": 16.4851,
+      "step": 9226
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.719429845332283e-05,
+      "loss": 16.6361,
+      "step": 9227
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.719129551854548e-05,
+      "loss": 15.5058,
+      "step": 9228
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.718829228343848e-05,
+      "loss": 16.0769,
+      "step": 9229
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.718528874802609e-05,
+      "loss": 16.7822,
+      "step": 9230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.718228491233256e-05,
+      "loss": 17.0119,
+      "step": 9231
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.717928077638214e-05,
+      "loss": 16.0965,
+      "step": 9232
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.71762763401991e-05,
+      "loss": 15.9019,
+      "step": 9233
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.717327160380771e-05,
+      "loss": 15.8628,
+      "step": 9234
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.717026656723219e-05,
+      "loss": 17.4833,
+      "step": 9235
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.716726123049688e-05,
+      "loss": 15.468,
+      "step": 9236
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.716425559362597e-05,
+      "loss": 16.935,
+      "step": 9237
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.716124965664379e-05,
+      "loss": 16.2473,
+      "step": 9238
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.71582434195746e-05,
+      "loss": 18.569,
+      "step": 9239
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.715523688244265e-05,
+      "loss": 19.3928,
+      "step": 9240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.715223004527224e-05,
+      "loss": 16.6299,
+      "step": 9241
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.714922290808766e-05,
+      "loss": 15.0381,
+      "step": 9242
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.714621547091317e-05,
+      "loss": 14.4718,
+      "step": 9243
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.714320773377307e-05,
+      "loss": 17.6519,
+      "step": 9244
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.714019969669165e-05,
+      "loss": 15.84,
+      "step": 9245
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.71371913596932e-05,
+      "loss": 17.2004,
+      "step": 9246
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.7134182722802e-05,
+      "loss": 15.5014,
+      "step": 9247
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.713117378604236e-05,
+      "loss": 15.0189,
+      "step": 9248
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.712816454943858e-05,
+      "loss": 16.7333,
+      "step": 9249
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.712515501301493e-05,
+      "loss": 14.7201,
+      "step": 9250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.712214517679575e-05,
+      "loss": 16.1003,
+      "step": 9251
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.711913504080534e-05,
+      "loss": 15.1988,
+      "step": 9252
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.7116124605068e-05,
+      "loss": 18.1469,
+      "step": 9253
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.711311386960803e-05,
+      "loss": 18.2265,
+      "step": 9254
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.711010283444976e-05,
+      "loss": 15.9754,
+      "step": 9255
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.710709149961748e-05,
+      "loss": 15.6306,
+      "step": 9256
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.710407986513554e-05,
+      "loss": 16.1425,
+      "step": 9257
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.710106793102824e-05,
+      "loss": 15.788,
+      "step": 9258
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.70980556973199e-05,
+      "loss": 14.1993,
+      "step": 9259
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.709504316403487e-05,
+      "loss": 17.0797,
+      "step": 9260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.709203033119743e-05,
+      "loss": 15.4034,
+      "step": 9261
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.708901719883195e-05,
+      "loss": 14.9502,
+      "step": 9262
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.708600376696276e-05,
+      "loss": 15.7481,
+      "step": 9263
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.708299003561418e-05,
+      "loss": 16.5544,
+      "step": 9264
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.707997600481052e-05,
+      "loss": 17.3915,
+      "step": 9265
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.707696167457619e-05,
+      "loss": 16.454,
+      "step": 9266
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.707394704493547e-05,
+      "loss": 16.7849,
+      "step": 9267
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.707093211591273e-05,
+      "loss": 16.2086,
+      "step": 9268
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.706791688753231e-05,
+      "loss": 15.8503,
+      "step": 9269
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.706490135981855e-05,
+      "loss": 15.9692,
+      "step": 9270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.706188553279581e-05,
+      "loss": 14.7169,
+      "step": 9271
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.705886940648846e-05,
+      "loss": 16.2891,
+      "step": 9272
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.705585298092083e-05,
+      "loss": 17.0944,
+      "step": 9273
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.705283625611728e-05,
+      "loss": 16.5978,
+      "step": 9274
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.70498192321022e-05,
+      "loss": 14.9654,
+      "step": 9275
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.704680190889994e-05,
+      "loss": 14.6411,
+      "step": 9276
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.704378428653483e-05,
+      "loss": 16.1012,
+      "step": 9277
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.704076636503128e-05,
+      "loss": 15.7124,
+      "step": 9278
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.703774814441365e-05,
+      "loss": 15.8692,
+      "step": 9279
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.70347296247063e-05,
+      "loss": 16.9679,
+      "step": 9280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.703171080593362e-05,
+      "loss": 16.6173,
+      "step": 9281
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.702869168812e-05,
+      "loss": 15.6102,
+      "step": 9282
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.702567227128977e-05,
+      "loss": 14.4474,
+      "step": 9283
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.702265255546738e-05,
+      "loss": 15.9071,
+      "step": 9284
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.701963254067716e-05,
+      "loss": 17.2052,
+      "step": 9285
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.701661222694353e-05,
+      "loss": 16.2674,
+      "step": 9286
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.701359161429086e-05,
+      "loss": 15.918,
+      "step": 9287
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.701057070274355e-05,
+      "loss": 14.8792,
+      "step": 9288
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.7007549492326e-05,
+      "loss": 17.2883,
+      "step": 9289
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.70045279830626e-05,
+      "loss": 15.6679,
+      "step": 9290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.700150617497775e-05,
+      "loss": 15.4075,
+      "step": 9291
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.699848406809585e-05,
+      "loss": 18.2515,
+      "step": 9292
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.699546166244132e-05,
+      "loss": 14.9458,
+      "step": 9293
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.699243895803856e-05,
+      "loss": 15.3544,
+      "step": 9294
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.698941595491196e-05,
+      "loss": 17.0034,
+      "step": 9295
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.698639265308595e-05,
+      "loss": 15.9944,
+      "step": 9296
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.698336905258492e-05,
+      "loss": 14.8015,
+      "step": 9297
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.698034515343334e-05,
+      "loss": 19.834,
+      "step": 9298
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.697732095565558e-05,
+      "loss": 15.2905,
+      "step": 9299
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.697429645927608e-05,
+      "loss": 16.4805,
+      "step": 9300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.697127166431926e-05,
+      "loss": 14.3991,
+      "step": 9301
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.696824657080953e-05,
+      "loss": 16.2508,
+      "step": 9302
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.696522117877135e-05,
+      "loss": 16.4721,
+      "step": 9303
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.696219548822913e-05,
+      "loss": 17.478,
+      "step": 9304
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.69591694992073e-05,
+      "loss": 18.1664,
+      "step": 9305
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.695614321173033e-05,
+      "loss": 15.5089,
+      "step": 9306
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.695311662582261e-05,
+      "loss": 18.0781,
+      "step": 9307
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.69500897415086e-05,
+      "loss": 17.1676,
+      "step": 9308
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.694706255881275e-05,
+      "loss": 17.1653,
+      "step": 9309
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8.69440350777595e-05,
+      "loss": 16.4685,
+      "step": 9310
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.69410072983733e-05,
+      "loss": 15.9255,
+      "step": 9311
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.69379792206786e-05,
+      "loss": 14.2145,
+      "step": 9312
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.693495084469984e-05,
+      "loss": 14.7642,
+      "step": 9313
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.69319221704615e-05,
+      "loss": 16.1186,
+      "step": 9314
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.692889319798803e-05,
+      "loss": 15.4786,
+      "step": 9315
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.692586392730387e-05,
+      "loss": 14.8553,
+      "step": 9316
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.692283435843349e-05,
+      "loss": 18.4576,
+      "step": 9317
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.691980449140135e-05,
+      "loss": 16.366,
+      "step": 9318
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.691677432623195e-05,
+      "loss": 16.3197,
+      "step": 9319
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.691374386294974e-05,
+      "loss": 16.3839,
+      "step": 9320
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.691071310157916e-05,
+      "loss": 13.7834,
+      "step": 9321
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.690768204214474e-05,
+      "loss": 16.6225,
+      "step": 9322
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.690465068467092e-05,
+      "loss": 15.9704,
+      "step": 9323
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.690161902918219e-05,
+      "loss": 15.9276,
+      "step": 9324
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.689858707570301e-05,
+      "loss": 18.3796,
+      "step": 9325
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.689555482425791e-05,
+      "loss": 15.7145,
+      "step": 9326
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.689252227487133e-05,
+      "loss": 17.3308,
+      "step": 9327
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.688948942756778e-05,
+      "loss": 15.8536,
+      "step": 9328
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.688645628237177e-05,
+      "loss": 15.9442,
+      "step": 9329
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.688342283930775e-05,
+      "loss": 15.3764,
+      "step": 9330
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.688038909840024e-05,
+      "loss": 15.2507,
+      "step": 9331
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.687735505967373e-05,
+      "loss": 14.5491,
+      "step": 9332
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.687432072315276e-05,
+      "loss": 16.4476,
+      "step": 9333
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.687128608886176e-05,
+      "loss": 15.9929,
+      "step": 9334
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.686825115682531e-05,
+      "loss": 16.8225,
+      "step": 9335
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.686521592706787e-05,
+      "loss": 17.1287,
+      "step": 9336
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.686218039961396e-05,
+      "loss": 15.1234,
+      "step": 9337
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.68591445744881e-05,
+      "loss": 16.366,
+      "step": 9338
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.68561084517148e-05,
+      "loss": 17.7483,
+      "step": 9339
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.685307203131858e-05,
+      "loss": 15.9577,
+      "step": 9340
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.685003531332396e-05,
+      "loss": 14.8129,
+      "step": 9341
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.684699829775544e-05,
+      "loss": 16.1547,
+      "step": 9342
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.684396098463759e-05,
+      "loss": 14.6942,
+      "step": 9343
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.684092337399492e-05,
+      "loss": 17.3991,
+      "step": 9344
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.683788546585195e-05,
+      "loss": 16.2998,
+      "step": 9345
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.683484726023319e-05,
+      "loss": 15.0894,
+      "step": 9346
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.683180875716322e-05,
+      "loss": 17.0165,
+      "step": 9347
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.682876995666656e-05,
+      "loss": 16.9247,
+      "step": 9348
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.682573085876773e-05,
+      "loss": 14.6971,
+      "step": 9349
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.682269146349129e-05,
+      "loss": 18.6329,
+      "step": 9350
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.681965177086179e-05,
+      "loss": 15.0429,
+      "step": 9351
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.681661178090376e-05,
+      "loss": 17.4513,
+      "step": 9352
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.681357149364175e-05,
+      "loss": 16.6578,
+      "step": 9353
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.681053090910032e-05,
+      "loss": 16.4804,
+      "step": 9354
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.680749002730402e-05,
+      "loss": 14.558,
+      "step": 9355
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.680444884827741e-05,
+      "loss": 16.1746,
+      "step": 9356
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.680140737204506e-05,
+      "loss": 16.1428,
+      "step": 9357
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.679836559863148e-05,
+      "loss": 16.1483,
+      "step": 9358
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.679532352806129e-05,
+      "loss": 15.9378,
+      "step": 9359
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.679228116035902e-05,
+      "loss": 15.8542,
+      "step": 9360
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.678923849554928e-05,
+      "loss": 15.9299,
+      "step": 9361
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.678619553365659e-05,
+      "loss": 16.5471,
+      "step": 9362
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.678315227470555e-05,
+      "loss": 16.5924,
+      "step": 9363
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.678010871872072e-05,
+      "loss": 16.2003,
+      "step": 9364
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.67770648657267e-05,
+      "loss": 16.1932,
+      "step": 9365
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.677402071574805e-05,
+      "loss": 16.2264,
+      "step": 9366
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.677097626880936e-05,
+      "loss": 17.5834,
+      "step": 9367
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.676793152493522e-05,
+      "loss": 15.8393,
+      "step": 9368
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.67648864841502e-05,
+      "loss": 15.6663,
+      "step": 9369
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.676184114647892e-05,
+      "loss": 15.1659,
+      "step": 9370
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.675879551194595e-05,
+      "loss": 15.2623,
+      "step": 9371
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.675574958057586e-05,
+      "loss": 18.6785,
+      "step": 9372
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.67527033523933e-05,
+      "loss": 17.3856,
+      "step": 9373
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.674965682742285e-05,
+      "loss": 16.2284,
+      "step": 9374
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.674661000568908e-05,
+      "loss": 14.4541,
+      "step": 9375
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.674356288721663e-05,
+      "loss": 17.8702,
+      "step": 9376
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.674051547203009e-05,
+      "loss": 14.733,
+      "step": 9377
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.673746776015408e-05,
+      "loss": 14.2238,
+      "step": 9378
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.67344197516132e-05,
+      "loss": 15.1791,
+      "step": 9379
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.673137144643206e-05,
+      "loss": 16.7983,
+      "step": 9380
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.67283228446353e-05,
+      "loss": 18.0248,
+      "step": 9381
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.672527394624752e-05,
+      "loss": 15.5605,
+      "step": 9382
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.672222475129334e-05,
+      "loss": 16.3207,
+      "step": 9383
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.671917525979738e-05,
+      "loss": 16.7148,
+      "step": 9384
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.671612547178428e-05,
+      "loss": 15.8781,
+      "step": 9385
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.671307538727866e-05,
+      "loss": 17.1189,
+      "step": 9386
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.671002500630513e-05,
+      "loss": 15.9059,
+      "step": 9387
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.670697432888837e-05,
+      "loss": 16.0136,
+      "step": 9388
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.670392335505298e-05,
+      "loss": 15.2087,
+      "step": 9389
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.670087208482359e-05,
+      "loss": 14.7345,
+      "step": 9390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.669782051822486e-05,
+      "loss": 17.18,
+      "step": 9391
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.669476865528143e-05,
+      "loss": 15.6846,
+      "step": 9392
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.669171649601795e-05,
+      "loss": 17.1319,
+      "step": 9393
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.668866404045905e-05,
+      "loss": 14.4128,
+      "step": 9394
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.668561128862938e-05,
+      "loss": 16.4006,
+      "step": 9395
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.66825582405536e-05,
+      "loss": 16.7151,
+      "step": 9396
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.667950489625638e-05,
+      "loss": 16.0843,
+      "step": 9397
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.667645125576235e-05,
+      "loss": 15.8807,
+      "step": 9398
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.667339731909618e-05,
+      "loss": 15.7618,
+      "step": 9399
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.667034308628252e-05,
+      "loss": 16.2308,
+      "step": 9400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.666728855734604e-05,
+      "loss": 16.061,
+      "step": 9401
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.666423373231144e-05,
+      "loss": 14.1735,
+      "step": 9402
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.666117861120334e-05,
+      "loss": 15.0669,
+      "step": 9403
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.665812319404643e-05,
+      "loss": 15.1582,
+      "step": 9404
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.66550674808654e-05,
+      "loss": 16.4061,
+      "step": 9405
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.665201147168489e-05,
+      "loss": 16.3434,
+      "step": 9406
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.66489551665296e-05,
+      "loss": 15.4991,
+      "step": 9407
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.66458985654242e-05,
+      "loss": 16.8955,
+      "step": 9408
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.664284166839339e-05,
+      "loss": 16.809,
+      "step": 9409
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.663978447546185e-05,
+      "loss": 16.1289,
+      "step": 9410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.663672698665425e-05,
+      "loss": 16.0186,
+      "step": 9411
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.66336692019953e-05,
+      "loss": 15.0657,
+      "step": 9412
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.66306111215097e-05,
+      "loss": 14.6781,
+      "step": 9413
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.662755274522211e-05,
+      "loss": 17.7164,
+      "step": 9414
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.662449407315725e-05,
+      "loss": 17.1036,
+      "step": 9415
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.662143510533982e-05,
+      "loss": 14.8476,
+      "step": 9416
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.661837584179453e-05,
+      "loss": 16.2386,
+      "step": 9417
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.661531628254606e-05,
+      "loss": 16.5907,
+      "step": 9418
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.661225642761915e-05,
+      "loss": 16.1428,
+      "step": 9419
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.660919627703846e-05,
+      "loss": 14.8943,
+      "step": 9420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.660613583082875e-05,
+      "loss": 15.8697,
+      "step": 9421
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.66030750890147e-05,
+      "loss": 16.2041,
+      "step": 9422
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.660001405162106e-05,
+      "loss": 16.3726,
+      "step": 9423
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.659695271867252e-05,
+      "loss": 15.7559,
+      "step": 9424
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.65938910901938e-05,
+      "loss": 18.2471,
+      "step": 9425
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.659082916620965e-05,
+      "loss": 16.8896,
+      "step": 9426
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.658776694674478e-05,
+      "loss": 18.1398,
+      "step": 9427
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.658470443182392e-05,
+      "loss": 14.5347,
+      "step": 9428
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.65816416214718e-05,
+      "loss": 16.4679,
+      "step": 9429
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.657857851571314e-05,
+      "loss": 15.7674,
+      "step": 9430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.657551511457269e-05,
+      "loss": 17.0461,
+      "step": 9431
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.65724514180752e-05,
+      "loss": 15.0462,
+      "step": 9432
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.656938742624539e-05,
+      "loss": 16.4776,
+      "step": 9433
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.6566323139108e-05,
+      "loss": 16.0987,
+      "step": 9434
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.656325855668778e-05,
+      "loss": 16.2559,
+      "step": 9435
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.656019367900949e-05,
+      "loss": 17.7003,
+      "step": 9436
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.655712850609787e-05,
+      "loss": 16.7549,
+      "step": 9437
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.655406303797767e-05,
+      "loss": 13.9545,
+      "step": 9438
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.655099727467364e-05,
+      "loss": 15.8201,
+      "step": 9439
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.654793121621054e-05,
+      "loss": 15.7369,
+      "step": 9440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.654486486261316e-05,
+      "loss": 16.1909,
+      "step": 9441
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.654179821390621e-05,
+      "loss": 16.9385,
+      "step": 9442
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.65387312701145e-05,
+      "loss": 14.6664,
+      "step": 9443
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.653566403126274e-05,
+      "loss": 17.2482,
+      "step": 9444
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.653259649737575e-05,
+      "loss": 13.984,
+      "step": 9445
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.65295286684783e-05,
+      "loss": 14.9777,
+      "step": 9446
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.652646054459514e-05,
+      "loss": 17.5494,
+      "step": 9447
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.652339212575104e-05,
+      "loss": 15.4446,
+      "step": 9448
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.65203234119708e-05,
+      "loss": 18.7251,
+      "step": 9449
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.65172544032792e-05,
+      "loss": 15.1532,
+      "step": 9450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.651418509970102e-05,
+      "loss": 14.4001,
+      "step": 9451
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.651111550126103e-05,
+      "loss": 16.7507,
+      "step": 9452
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.650804560798403e-05,
+      "loss": 14.5072,
+      "step": 9453
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.650497541989482e-05,
+      "loss": 16.7918,
+      "step": 9454
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.650190493701817e-05,
+      "loss": 14.8933,
+      "step": 9455
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.64988341593789e-05,
+      "loss": 15.4588,
+      "step": 9456
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.649576308700178e-05,
+      "loss": 14.5479,
+      "step": 9457
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.649269171991164e-05,
+      "loss": 16.4656,
+      "step": 9458
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.648962005813326e-05,
+      "loss": 15.9911,
+      "step": 9459
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.648654810169143e-05,
+      "loss": 14.8516,
+      "step": 9460
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.648347585061101e-05,
+      "loss": 17.6995,
+      "step": 9461
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.648040330491675e-05,
+      "loss": 16.0882,
+      "step": 9462
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.647733046463351e-05,
+      "loss": 15.0308,
+      "step": 9463
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.647425732978607e-05,
+      "loss": 15.867,
+      "step": 9464
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.647118390039928e-05,
+      "loss": 16.8393,
+      "step": 9465
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.646811017649792e-05,
+      "loss": 15.7667,
+      "step": 9466
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.646503615810683e-05,
+      "loss": 16.6998,
+      "step": 9467
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.646196184525082e-05,
+      "loss": 18.0978,
+      "step": 9468
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.645888723795475e-05,
+      "loss": 16.2201,
+      "step": 9469
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.645581233624342e-05,
+      "loss": 14.8511,
+      "step": 9470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.645273714014166e-05,
+      "loss": 16.2997,
+      "step": 9471
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.644966164967432e-05,
+      "loss": 15.8516,
+      "step": 9472
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.644658586486623e-05,
+      "loss": 15.3546,
+      "step": 9473
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.644350978574221e-05,
+      "loss": 15.3894,
+      "step": 9474
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.644043341232711e-05,
+      "loss": 16.0817,
+      "step": 9475
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.643735674464579e-05,
+      "loss": 14.642,
+      "step": 9476
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.643427978272306e-05,
+      "loss": 17.3634,
+      "step": 9477
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.643120252658381e-05,
+      "loss": 15.8551,
+      "step": 9478
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.642812497625288e-05,
+      "loss": 15.5202,
+      "step": 9479
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.642504713175508e-05,
+      "loss": 15.4844,
+      "step": 9480
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.642196899311533e-05,
+      "loss": 19.798,
+      "step": 9481
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.64188905603584e-05,
+      "loss": 16.0403,
+      "step": 9482
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.641581183350924e-05,
+      "loss": 15.3696,
+      "step": 9483
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.641273281259267e-05,
+      "loss": 17.6318,
+      "step": 9484
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.640965349763355e-05,
+      "loss": 14.076,
+      "step": 9485
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.640657388865674e-05,
+      "loss": 17.6449,
+      "step": 9486
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.640349398568713e-05,
+      "loss": 16.0624,
+      "step": 9487
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.640041378874959e-05,
+      "loss": 17.486,
+      "step": 9488
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.639733329786898e-05,
+      "loss": 15.2518,
+      "step": 9489
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.639425251307019e-05,
+      "loss": 15.9868,
+      "step": 9490
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.639117143437807e-05,
+      "loss": 15.8677,
+      "step": 9491
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.638809006181753e-05,
+      "loss": 14.8077,
+      "step": 9492
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.638500839541344e-05,
+      "loss": 15.9249,
+      "step": 9493
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.638192643519069e-05,
+      "loss": 16.5472,
+      "step": 9494
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.637884418117416e-05,
+      "loss": 16.3297,
+      "step": 9495
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.637576163338873e-05,
+      "loss": 15.1255,
+      "step": 9496
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.637267879185936e-05,
+      "loss": 17.297,
+      "step": 9497
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.636959565661084e-05,
+      "loss": 16.6491,
+      "step": 9498
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.636651222766815e-05,
+      "loss": 14.9152,
+      "step": 9499
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.636342850505616e-05,
+      "loss": 16.7432,
+      "step": 9500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.636034448879974e-05,
+      "loss": 18.3559,
+      "step": 9501
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.635726017892387e-05,
+      "loss": 15.8197,
+      "step": 9502
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.635417557545338e-05,
+      "loss": 17.6301,
+      "step": 9503
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.635109067841322e-05,
+      "loss": 15.0611,
+      "step": 9504
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.63480054878283e-05,
+      "loss": 15.494,
+      "step": 9505
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.634492000372351e-05,
+      "loss": 16.6932,
+      "step": 9506
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.634183422612381e-05,
+      "loss": 15.4742,
+      "step": 9507
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.633874815505408e-05,
+      "loss": 16.2562,
+      "step": 9508
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.633566179053923e-05,
+      "loss": 16.0025,
+      "step": 9509
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.633257513260422e-05,
+      "loss": 14.1328,
+      "step": 9510
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.632948818127398e-05,
+      "loss": 15.2007,
+      "step": 9511
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.63264009365734e-05,
+      "loss": 18.0108,
+      "step": 9512
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.632331339852742e-05,
+      "loss": 16.5665,
+      "step": 9513
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.632022556716099e-05,
+      "loss": 17.0788,
+      "step": 9514
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.631713744249904e-05,
+      "loss": 15.6972,
+      "step": 9515
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.63140490245665e-05,
+      "loss": 16.9218,
+      "step": 9516
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.631096031338832e-05,
+      "loss": 15.5684,
+      "step": 9517
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.630787130898943e-05,
+      "loss": 15.2369,
+      "step": 9518
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.630478201139479e-05,
+      "loss": 16.6538,
+      "step": 9519
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.630169242062932e-05,
+      "loss": 15.4204,
+      "step": 9520
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.6298602536718e-05,
+      "loss": 16.4404,
+      "step": 9521
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.629551235968577e-05,
+      "loss": 15.0332,
+      "step": 9522
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.629242188955757e-05,
+      "loss": 16.3719,
+      "step": 9523
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.628933112635837e-05,
+      "loss": 14.9696,
+      "step": 9524
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.628624007011314e-05,
+      "loss": 15.3673,
+      "step": 9525
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.628314872084684e-05,
+      "loss": 15.5605,
+      "step": 9526
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.628005707858439e-05,
+      "loss": 16.1957,
+      "step": 9527
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.62769651433508e-05,
+      "loss": 16.5545,
+      "step": 9528
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.627387291517103e-05,
+      "loss": 15.5335,
+      "step": 9529
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.627078039407004e-05,
+      "loss": 17.2469,
+      "step": 9530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.626768758007283e-05,
+      "loss": 17.6709,
+      "step": 9531
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.626459447320435e-05,
+      "loss": 17.3512,
+      "step": 9532
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.626150107348956e-05,
+      "loss": 18.7954,
+      "step": 9533
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.625840738095347e-05,
+      "loss": 15.6569,
+      "step": 9534
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.625531339562108e-05,
+      "loss": 17.2066,
+      "step": 9535
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.625221911751732e-05,
+      "loss": 18.0846,
+      "step": 9536
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.624912454666723e-05,
+      "loss": 16.4308,
+      "step": 9537
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.624602968309576e-05,
+      "loss": 17.8965,
+      "step": 9538
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.624293452682792e-05,
+      "loss": 14.9998,
+      "step": 9539
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.62398390778887e-05,
+      "loss": 14.7565,
+      "step": 9540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.623674333630309e-05,
+      "loss": 15.9878,
+      "step": 9541
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.623364730209612e-05,
+      "loss": 16.3422,
+      "step": 9542
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.623055097529275e-05,
+      "loss": 16.1888,
+      "step": 9543
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.622745435591803e-05,
+      "loss": 18.4727,
+      "step": 9544
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.622435744399693e-05,
+      "loss": 18.1731,
+      "step": 9545
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.622126023955446e-05,
+      "loss": 15.928,
+      "step": 9546
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.621816274261563e-05,
+      "loss": 15.8488,
+      "step": 9547
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.621506495320547e-05,
+      "loss": 14.8221,
+      "step": 9548
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.621196687134898e-05,
+      "loss": 15.3433,
+      "step": 9549
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.620886849707119e-05,
+      "loss": 19.7788,
+      "step": 9550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.620576983039712e-05,
+      "loss": 16.5762,
+      "step": 9551
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.620267087135178e-05,
+      "loss": 17.1267,
+      "step": 9552
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.619957161996022e-05,
+      "loss": 16.6587,
+      "step": 9553
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.619647207624742e-05,
+      "loss": 17.1156,
+      "step": 9554
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.619337224023846e-05,
+      "loss": 16.1576,
+      "step": 9555
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.619027211195836e-05,
+      "loss": 15.473,
+      "step": 9556
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.618717169143213e-05,
+      "loss": 15.7643,
+      "step": 9557
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.618407097868482e-05,
+      "loss": 16.2846,
+      "step": 9558
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.618096997374148e-05,
+      "loss": 15.0941,
+      "step": 9559
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.617786867662713e-05,
+      "loss": 17.2889,
+      "step": 9560
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.617476708736683e-05,
+      "loss": 15.0357,
+      "step": 9561
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.617166520598563e-05,
+      "loss": 16.7105,
+      "step": 9562
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.616856303250856e-05,
+      "loss": 16.3386,
+      "step": 9563
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.616546056696068e-05,
+      "loss": 16.7584,
+      "step": 9564
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.616235780936706e-05,
+      "loss": 15.6904,
+      "step": 9565
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.615925475975275e-05,
+      "loss": 16.3205,
+      "step": 9566
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.615615141814279e-05,
+      "loss": 16.9821,
+      "step": 9567
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.615304778456224e-05,
+      "loss": 16.6922,
+      "step": 9568
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.614994385903618e-05,
+      "loss": 16.2166,
+      "step": 9569
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.614683964158967e-05,
+      "loss": 15.722,
+      "step": 9570
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.614373513224776e-05,
+      "loss": 16.9786,
+      "step": 9571
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.614063033103555e-05,
+      "loss": 15.3916,
+      "step": 9572
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.613752523797808e-05,
+      "loss": 14.9514,
+      "step": 9573
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.613441985310045e-05,
+      "loss": 15.4318,
+      "step": 9574
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.613131417642772e-05,
+      "loss": 17.2065,
+      "step": 9575
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.612820820798499e-05,
+      "loss": 16.6406,
+      "step": 9576
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.612510194779732e-05,
+      "loss": 14.8306,
+      "step": 9577
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.612199539588982e-05,
+      "loss": 16.5553,
+      "step": 9578
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.611888855228753e-05,
+      "loss": 15.6234,
+      "step": 9579
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.611578141701558e-05,
+      "loss": 15.3328,
+      "step": 9580
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.611267399009903e-05,
+      "loss": 17.7596,
+      "step": 9581
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.610956627156302e-05,
+      "loss": 17.4096,
+      "step": 9582
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.61064582614326e-05,
+      "loss": 16.5589,
+      "step": 9583
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.610334995973289e-05,
+      "loss": 15.8492,
+      "step": 9584
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.610024136648899e-05,
+      "loss": 16.8936,
+      "step": 9585
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.609713248172599e-05,
+      "loss": 16.1768,
+      "step": 9586
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.609402330546901e-05,
+      "loss": 17.7678,
+      "step": 9587
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.609091383774313e-05,
+      "loss": 16.8707,
+      "step": 9588
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.60878040785735e-05,
+      "loss": 15.428,
+      "step": 9589
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.60846940279852e-05,
+      "loss": 14.5624,
+      "step": 9590
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.608158368600337e-05,
+      "loss": 15.1919,
+      "step": 9591
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.60784730526531e-05,
+      "loss": 17.1319,
+      "step": 9592
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.607536212795954e-05,
+      "loss": 17.1584,
+      "step": 9593
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.607225091194779e-05,
+      "loss": 16.0622,
+      "step": 9594
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.606913940464297e-05,
+      "loss": 15.8743,
+      "step": 9595
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.606602760607022e-05,
+      "loss": 15.4668,
+      "step": 9596
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.606291551625466e-05,
+      "loss": 17.2908,
+      "step": 9597
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.605980313522142e-05,
+      "loss": 15.909,
+      "step": 9598
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.605669046299565e-05,
+      "loss": 16.2483,
+      "step": 9599
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.605357749960247e-05,
+      "loss": 15.0265,
+      "step": 9600
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.605046424506702e-05,
+      "loss": 15.4658,
+      "step": 9601
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.604735069941443e-05,
+      "loss": 15.7191,
+      "step": 9602
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.604423686266986e-05,
+      "loss": 16.6955,
+      "step": 9603
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.604112273485846e-05,
+      "loss": 17.039,
+      "step": 9604
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.603800831600536e-05,
+      "loss": 14.7747,
+      "step": 9605
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.603489360613571e-05,
+      "loss": 16.3411,
+      "step": 9606
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.603177860527467e-05,
+      "loss": 15.0358,
+      "step": 9607
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.60286633134474e-05,
+      "loss": 14.5967,
+      "step": 9608
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.602554773067904e-05,
+      "loss": 15.2678,
+      "step": 9609
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.602243185699477e-05,
+      "loss": 15.6768,
+      "step": 9610
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.601931569241973e-05,
+      "loss": 17.4462,
+      "step": 9611
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.60161992369791e-05,
+      "loss": 14.7904,
+      "step": 9612
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.601308249069803e-05,
+      "loss": 16.7715,
+      "step": 9613
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.600996545360172e-05,
+      "loss": 15.1085,
+      "step": 9614
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.600684812571532e-05,
+      "loss": 15.8182,
+      "step": 9615
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.600373050706399e-05,
+      "loss": 15.5817,
+      "step": 9616
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.600061259767291e-05,
+      "loss": 15.4746,
+      "step": 9617
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.599749439756728e-05,
+      "loss": 16.1549,
+      "step": 9618
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.599437590677228e-05,
+      "loss": 16.6821,
+      "step": 9619
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.599125712531306e-05,
+      "loss": 15.5249,
+      "step": 9620
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.598813805321483e-05,
+      "loss": 15.2562,
+      "step": 9621
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.598501869050277e-05,
+      "loss": 17.3968,
+      "step": 9622
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.598189903720207e-05,
+      "loss": 17.3395,
+      "step": 9623
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.597877909333794e-05,
+      "loss": 16.4893,
+      "step": 9624
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.597565885893554e-05,
+      "loss": 15.4229,
+      "step": 9625
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.597253833402008e-05,
+      "loss": 15.5352,
+      "step": 9626
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.596941751861678e-05,
+      "loss": 16.2962,
+      "step": 9627
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.596629641275083e-05,
+      "loss": 17.8774,
+      "step": 9628
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.596317501644741e-05,
+      "loss": 16.3773,
+      "step": 9629
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.596005332973175e-05,
+      "loss": 15.4248,
+      "step": 9630
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.595693135262905e-05,
+      "loss": 16.5595,
+      "step": 9631
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.595380908516454e-05,
+      "loss": 15.139,
+      "step": 9632
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.59506865273634e-05,
+      "loss": 16.0853,
+      "step": 9633
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.594756367925088e-05,
+      "loss": 17.9115,
+      "step": 9634
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.594444054085217e-05,
+      "loss": 13.6501,
+      "step": 9635
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.594131711219251e-05,
+      "loss": 16.1505,
+      "step": 9636
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.593819339329712e-05,
+      "loss": 15.9453,
+      "step": 9637
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.59350693841912e-05,
+      "loss": 14.9919,
+      "step": 9638
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.59319450849e-05,
+      "loss": 17.2359,
+      "step": 9639
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.592882049544875e-05,
+      "loss": 15.2957,
+      "step": 9640
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.592569561586268e-05,
+      "loss": 15.8264,
+      "step": 9641
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.592257044616702e-05,
+      "loss": 16.1689,
+      "step": 9642
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.5919444986387e-05,
+      "loss": 16.5366,
+      "step": 9643
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.591631923654788e-05,
+      "loss": 14.2487,
+      "step": 9644
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.591319319667487e-05,
+      "loss": 17.2181,
+      "step": 9645
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.591006686679324e-05,
+      "loss": 17.0271,
+      "step": 9646
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.590694024692824e-05,
+      "loss": 14.9361,
+      "step": 9647
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.590381333710508e-05,
+      "loss": 15.8699,
+      "step": 9648
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.590068613734906e-05,
+      "loss": 15.6003,
+      "step": 9649
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.589755864768541e-05,
+      "loss": 16.1651,
+      "step": 9650
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.589443086813937e-05,
+      "loss": 15.4575,
+      "step": 9651
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.589130279873623e-05,
+      "loss": 15.5074,
+      "step": 9652
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.588817443950122e-05,
+      "loss": 18.9249,
+      "step": 9653
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.588504579045961e-05,
+      "loss": 15.3541,
+      "step": 9654
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.588191685163667e-05,
+      "loss": 17.9612,
+      "step": 9655
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.587878762305768e-05,
+      "loss": 15.9314,
+      "step": 9656
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.587565810474789e-05,
+      "loss": 14.6676,
+      "step": 9657
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.587252829673258e-05,
+      "loss": 17.3081,
+      "step": 9658
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.5869398199037e-05,
+      "loss": 16.8021,
+      "step": 9659
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.586626781168648e-05,
+      "loss": 13.9445,
+      "step": 9660
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.586313713470626e-05,
+      "loss": 15.3864,
+      "step": 9661
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.586000616812162e-05,
+      "loss": 15.9723,
+      "step": 9662
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.585687491195785e-05,
+      "loss": 17.5293,
+      "step": 9663
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.585374336624025e-05,
+      "loss": 16.4724,
+      "step": 9664
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.585061153099407e-05,
+      "loss": 16.8897,
+      "step": 9665
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.584747940624465e-05,
+      "loss": 16.1617,
+      "step": 9666
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.584434699201725e-05,
+      "loss": 15.5554,
+      "step": 9667
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.584121428833718e-05,
+      "loss": 17.286,
+      "step": 9668
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.583808129522972e-05,
+      "loss": 16.3706,
+      "step": 9669
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.583494801272018e-05,
+      "loss": 15.9667,
+      "step": 9670
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.583181444083385e-05,
+      "loss": 15.4975,
+      "step": 9671
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.582868057959608e-05,
+      "loss": 15.9932,
+      "step": 9672
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.582554642903212e-05,
+      "loss": 14.9344,
+      "step": 9673
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.582241198916732e-05,
+      "loss": 16.848,
+      "step": 9674
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.581927726002696e-05,
+      "loss": 16.7771,
+      "step": 9675
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.581614224163636e-05,
+      "loss": 14.0271,
+      "step": 9676
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.581300693402086e-05,
+      "loss": 14.6678,
+      "step": 9677
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.580987133720576e-05,
+      "loss": 15.9854,
+      "step": 9678
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.580673545121638e-05,
+      "loss": 14.7958,
+      "step": 9679
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.580359927607805e-05,
+      "loss": 14.6679,
+      "step": 9680
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.580046281181609e-05,
+      "loss": 17.6609,
+      "step": 9681
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.579732605845583e-05,
+      "loss": 14.6113,
+      "step": 9682
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.579418901602261e-05,
+      "loss": 16.3374,
+      "step": 9683
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.579105168454173e-05,
+      "loss": 17.6888,
+      "step": 9684
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.578791406403857e-05,
+      "loss": 16.0216,
+      "step": 9685
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.578477615453844e-05,
+      "loss": 16.0545,
+      "step": 9686
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.578163795606666e-05,
+      "loss": 16.2659,
+      "step": 9687
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.577849946864861e-05,
+      "loss": 18.84,
+      "step": 9688
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.577536069230964e-05,
+      "loss": 15.1773,
+      "step": 9689
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.577222162707505e-05,
+      "loss": 15.2392,
+      "step": 9690
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.576908227297023e-05,
+      "loss": 17.5517,
+      "step": 9691
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.576594263002051e-05,
+      "loss": 16.4803,
+      "step": 9692
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.576280269825125e-05,
+      "loss": 17.2182,
+      "step": 9693
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.575966247768781e-05,
+      "loss": 16.2079,
+      "step": 9694
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.575652196835552e-05,
+      "loss": 16.293,
+      "step": 9695
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.57533811702798e-05,
+      "loss": 16.6904,
+      "step": 9696
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.575024008348595e-05,
+      "loss": 15.0834,
+      "step": 9697
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.574709870799938e-05,
+      "loss": 16.7276,
+      "step": 9698
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.574395704384543e-05,
+      "loss": 16.6786,
+      "step": 9699
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.574081509104948e-05,
+      "loss": 17.4158,
+      "step": 9700
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.573767284963688e-05,
+      "loss": 15.8458,
+      "step": 9701
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.573453031963306e-05,
+      "loss": 16.4998,
+      "step": 9702
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.573138750106334e-05,
+      "loss": 16.7183,
+      "step": 9703
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.572824439395313e-05,
+      "loss": 16.4071,
+      "step": 9704
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.572510099832781e-05,
+      "loss": 14.1024,
+      "step": 9705
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.572195731421274e-05,
+      "loss": 14.9,
+      "step": 9706
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.571881334163334e-05,
+      "loss": 15.8107,
+      "step": 9707
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.571566908061497e-05,
+      "loss": 16.8162,
+      "step": 9708
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.571252453118302e-05,
+      "loss": 15.6696,
+      "step": 9709
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.570937969336292e-05,
+      "loss": 15.539,
+      "step": 9710
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.570623456718002e-05,
+      "loss": 15.0653,
+      "step": 9711
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.570308915265974e-05,
+      "loss": 15.7056,
+      "step": 9712
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.56999434498275e-05,
+      "loss": 15.883,
+      "step": 9713
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.569679745870866e-05,
+      "loss": 15.847,
+      "step": 9714
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.569365117932865e-05,
+      "loss": 17.0913,
+      "step": 9715
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.569050461171288e-05,
+      "loss": 16.5191,
+      "step": 9716
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.568735775588675e-05,
+      "loss": 16.3919,
+      "step": 9717
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.568421061187567e-05,
+      "loss": 14.8074,
+      "step": 9718
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.568106317970503e-05,
+      "loss": 14.7646,
+      "step": 9719
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.567791545940032e-05,
+      "loss": 15.7262,
+      "step": 9720
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.567476745098688e-05,
+      "loss": 15.9727,
+      "step": 9721
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.567161915449018e-05,
+      "loss": 16.2634,
+      "step": 9722
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.56684705699356e-05,
+      "loss": 16.5127,
+      "step": 9723
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.566532169734862e-05,
+      "loss": 15.1512,
+      "step": 9724
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.566217253675464e-05,
+      "loss": 14.2897,
+      "step": 9725
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.565902308817907e-05,
+      "loss": 15.0379,
+      "step": 9726
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.56558733516474e-05,
+      "loss": 14.7433,
+      "step": 9727
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.5652723327185e-05,
+      "loss": 16.4855,
+      "step": 9728
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.564957301481734e-05,
+      "loss": 16.2148,
+      "step": 9729
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.564642241456986e-05,
+      "loss": 16.1798,
+      "step": 9730
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.5643271526468e-05,
+      "loss": 15.4204,
+      "step": 9731
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.56401203505372e-05,
+      "loss": 15.5852,
+      "step": 9732
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.563696888680289e-05,
+      "loss": 14.1275,
+      "step": 9733
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.563381713529056e-05,
+      "loss": 17.8554,
+      "step": 9734
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.563066509602562e-05,
+      "loss": 15.6481,
+      "step": 9735
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.562751276903356e-05,
+      "loss": 15.1521,
+      "step": 9736
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.56243601543398e-05,
+      "loss": 15.117,
+      "step": 9737
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.562120725196983e-05,
+      "loss": 15.9874,
+      "step": 9738
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.561805406194908e-05,
+      "loss": 15.4199,
+      "step": 9739
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.561490058430304e-05,
+      "loss": 14.5478,
+      "step": 9740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.561174681905718e-05,
+      "loss": 15.0632,
+      "step": 9741
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.560859276623692e-05,
+      "loss": 15.1627,
+      "step": 9742
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.560543842586779e-05,
+      "loss": 16.499,
+      "step": 9743
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.560228379797522e-05,
+      "loss": 17.2981,
+      "step": 9744
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.559912888258468e-05,
+      "loss": 15.3467,
+      "step": 9745
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.559597367972168e-05,
+      "loss": 17.1032,
+      "step": 9746
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.559281818941169e-05,
+      "loss": 14.5706,
+      "step": 9747
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.558966241168019e-05,
+      "loss": 14.1516,
+      "step": 9748
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.558650634655265e-05,
+      "loss": 15.0406,
+      "step": 9749
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.558334999405457e-05,
+      "loss": 14.7691,
+      "step": 9750
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.558019335421143e-05,
+      "loss": 17.557,
+      "step": 9751
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.557703642704873e-05,
+      "loss": 16.6334,
+      "step": 9752
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.557387921259196e-05,
+      "loss": 15.3514,
+      "step": 9753
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.55707217108666e-05,
+      "loss": 16.5514,
+      "step": 9754
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.556756392189818e-05,
+      "loss": 16.1446,
+      "step": 9755
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.556440584571216e-05,
+      "loss": 16.8987,
+      "step": 9756
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.556124748233407e-05,
+      "loss": 15.2323,
+      "step": 9757
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.55580888317894e-05,
+      "loss": 15.071,
+      "step": 9758
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.555492989410367e-05,
+      "loss": 16.3835,
+      "step": 9759
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.55517706693024e-05,
+      "loss": 15.9944,
+      "step": 9760
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.554861115741106e-05,
+      "loss": 16.3736,
+      "step": 9761
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.55454513584552e-05,
+      "loss": 16.0963,
+      "step": 9762
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.554229127246034e-05,
+      "loss": 16.5347,
+      "step": 9763
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.553913089945196e-05,
+      "loss": 14.6231,
+      "step": 9764
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.553597023945562e-05,
+      "loss": 17.5382,
+      "step": 9765
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.553280929249682e-05,
+      "loss": 15.9521,
+      "step": 9766
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.55296480586011e-05,
+      "loss": 17.3102,
+      "step": 9767
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.552648653779399e-05,
+      "loss": 14.4937,
+      "step": 9768
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.552332473010098e-05,
+      "loss": 14.943,
+      "step": 9769
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.552016263554766e-05,
+      "loss": 16.1766,
+      "step": 9770
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.551700025415953e-05,
+      "loss": 15.7261,
+      "step": 9771
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.551383758596214e-05,
+      "loss": 16.0699,
+      "step": 9772
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.551067463098102e-05,
+      "loss": 15.3273,
+      "step": 9773
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.550751138924171e-05,
+      "loss": 15.5737,
+      "step": 9774
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.550434786076977e-05,
+      "loss": 15.5384,
+      "step": 9775
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.550118404559075e-05,
+      "loss": 16.5537,
+      "step": 9776
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.549801994373017e-05,
+      "loss": 14.4846,
+      "step": 9777
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.54948555552136e-05,
+      "loss": 16.1742,
+      "step": 9778
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.549169088006659e-05,
+      "loss": 15.8848,
+      "step": 9779
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.54885259183147e-05,
+      "loss": 17.5255,
+      "step": 9780
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.548536066998346e-05,
+      "loss": 13.3253,
+      "step": 9781
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.548219513509847e-05,
+      "loss": 15.8,
+      "step": 9782
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.547902931368528e-05,
+      "loss": 17.1491,
+      "step": 9783
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.547586320576945e-05,
+      "loss": 15.2762,
+      "step": 9784
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.547269681137655e-05,
+      "loss": 16.8009,
+      "step": 9785
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.546953013053214e-05,
+      "loss": 14.7144,
+      "step": 9786
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.54663631632618e-05,
+      "loss": 16.8202,
+      "step": 9787
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.546319590959113e-05,
+      "loss": 16.1975,
+      "step": 9788
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.546002836954566e-05,
+      "loss": 17.2672,
+      "step": 9789
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.545686054315097e-05,
+      "loss": 14.4085,
+      "step": 9790
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.545369243043269e-05,
+      "loss": 16.2293,
+      "step": 9791
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.545052403141635e-05,
+      "loss": 15.931,
+      "step": 9792
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.544735534612757e-05,
+      "loss": 18.7176,
+      "step": 9793
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.544418637459192e-05,
+      "loss": 15.288,
+      "step": 9794
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.5441017116835e-05,
+      "loss": 15.962,
+      "step": 9795
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.543784757288242e-05,
+      "loss": 16.5311,
+      "step": 9796
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.543467774275972e-05,
+      "loss": 15.3577,
+      "step": 9797
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.543150762649257e-05,
+      "loss": 15.8788,
+      "step": 9798
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.54283372241065e-05,
+      "loss": 15.918,
+      "step": 9799
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.542516653562716e-05,
+      "loss": 15.6501,
+      "step": 9800
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.542199556108013e-05,
+      "loss": 16.7843,
+      "step": 9801
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.541882430049102e-05,
+      "loss": 16.8159,
+      "step": 9802
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.541565275388546e-05,
+      "loss": 19.3089,
+      "step": 9803
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.541248092128903e-05,
+      "loss": 16.9684,
+      "step": 9804
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.540930880272736e-05,
+      "loss": 16.5893,
+      "step": 9805
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.540613639822605e-05,
+      "loss": 16.8252,
+      "step": 9806
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.540296370781076e-05,
+      "loss": 15.3264,
+      "step": 9807
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.539979073150706e-05,
+      "loss": 17.0468,
+      "step": 9808
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.53966174693406e-05,
+      "loss": 14.4997,
+      "step": 9809
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.539344392133699e-05,
+      "loss": 15.9435,
+      "step": 9810
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.539027008752188e-05,
+      "loss": 17.4562,
+      "step": 9811
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.538709596792088e-05,
+      "loss": 16.7041,
+      "step": 9812
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.538392156255963e-05,
+      "loss": 16.9445,
+      "step": 9813
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.538074687146376e-05,
+      "loss": 15.5129,
+      "step": 9814
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.53775718946589e-05,
+      "loss": 17.2496,
+      "step": 9815
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.537439663217071e-05,
+      "loss": 15.1427,
+      "step": 9816
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.53712210840248e-05,
+      "loss": 17.7631,
+      "step": 9817
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.536804525024684e-05,
+      "loss": 15.2586,
+      "step": 9818
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.536486913086249e-05,
+      "loss": 15.0748,
+      "step": 9819
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.536169272589734e-05,
+      "loss": 15.9936,
+      "step": 9820
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.53585160353771e-05,
+      "loss": 16.1108,
+      "step": 9821
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.535533905932738e-05,
+      "loss": 14.6225,
+      "step": 9822
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.535216179777386e-05,
+      "loss": 14.9268,
+      "step": 9823
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.53489842507422e-05,
+      "loss": 16.0742,
+      "step": 9824
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.534580641825803e-05,
+      "loss": 15.0897,
+      "step": 9825
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.534262830034705e-05,
+      "loss": 16.6615,
+      "step": 9826
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.53394498970349e-05,
+      "loss": 17.1107,
+      "step": 9827
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.533627120834724e-05,
+      "loss": 16.8678,
+      "step": 9828
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.533309223430977e-05,
+      "loss": 15.631,
+      "step": 9829
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.532991297494813e-05,
+      "loss": 16.6463,
+      "step": 9830
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.532673343028801e-05,
+      "loss": 13.2466,
+      "step": 9831
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.532355360035509e-05,
+      "loss": 14.8122,
+      "step": 9832
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.532037348517501e-05,
+      "loss": 16.3142,
+      "step": 9833
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.531719308477351e-05,
+      "loss": 14.4574,
+      "step": 9834
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.531401239917623e-05,
+      "loss": 16.2896,
+      "step": 9835
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.531083142840886e-05,
+      "loss": 16.7611,
+      "step": 9836
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.530765017249711e-05,
+      "loss": 16.8917,
+      "step": 9837
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.530446863146664e-05,
+      "loss": 14.8703,
+      "step": 9838
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.530128680534316e-05,
+      "loss": 17.3351,
+      "step": 9839
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.529810469415236e-05,
+      "loss": 14.8668,
+      "step": 9840
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.529492229791993e-05,
+      "loss": 18.4621,
+      "step": 9841
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.529173961667159e-05,
+      "loss": 15.9549,
+      "step": 9842
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.528855665043301e-05,
+      "loss": 15.0547,
+      "step": 9843
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.528537339922993e-05,
+      "loss": 16.4848,
+      "step": 9844
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.528218986308801e-05,
+      "loss": 16.2117,
+      "step": 9845
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.527900604203298e-05,
+      "loss": 16.0068,
+      "step": 9846
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.527582193609057e-05,
+      "loss": 16.6516,
+      "step": 9847
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.527263754528647e-05,
+      "loss": 17.382,
+      "step": 9848
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.52694528696464e-05,
+      "loss": 14.7473,
+      "step": 9849
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.526626790919607e-05,
+      "loss": 15.707,
+      "step": 9850
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.526308266396121e-05,
+      "loss": 15.952,
+      "step": 9851
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.525989713396754e-05,
+      "loss": 16.1876,
+      "step": 9852
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.525671131924079e-05,
+      "loss": 16.1075,
+      "step": 9853
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.525352521980666e-05,
+      "loss": 16.0347,
+      "step": 9854
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.525033883569091e-05,
+      "loss": 17.4791,
+      "step": 9855
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.524715216691925e-05,
+      "loss": 14.4213,
+      "step": 9856
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.524396521351743e-05,
+      "loss": 15.6515,
+      "step": 9857
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.524077797551116e-05,
+      "loss": 16.101,
+      "step": 9858
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.523759045292621e-05,
+      "loss": 14.8806,
+      "step": 9859
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.52344026457883e-05,
+      "loss": 15.2444,
+      "step": 9860
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.523121455412318e-05,
+      "loss": 15.1157,
+      "step": 9861
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.522802617795659e-05,
+      "loss": 16.9419,
+      "step": 9862
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.522483751731427e-05,
+      "loss": 14.5061,
+      "step": 9863
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.522164857222199e-05,
+      "loss": 14.0759,
+      "step": 9864
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.52184593427055e-05,
+      "loss": 16.1336,
+      "step": 9865
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.521526982879051e-05,
+      "loss": 16.0013,
+      "step": 9866
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.521208003050282e-05,
+      "loss": 16.2976,
+      "step": 9867
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.52088899478682e-05,
+      "loss": 16.7843,
+      "step": 9868
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.520569958091237e-05,
+      "loss": 15.2535,
+      "step": 9869
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.520250892966112e-05,
+      "loss": 17.9238,
+      "step": 9870
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.51993179941402e-05,
+      "loss": 17.334,
+      "step": 9871
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.519612677437538e-05,
+      "loss": 16.7272,
+      "step": 9872
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.519293527039244e-05,
+      "loss": 17.2166,
+      "step": 9873
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.518974348221716e-05,
+      "loss": 17.364,
+      "step": 9874
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.518655140987529e-05,
+      "loss": 15.5168,
+      "step": 9875
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.518335905339262e-05,
+      "loss": 17.7513,
+      "step": 9876
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.518016641279492e-05,
+      "loss": 16.0298,
+      "step": 9877
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.517697348810798e-05,
+      "loss": 16.4574,
+      "step": 9878
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.51737802793576e-05,
+      "loss": 16.4731,
+      "step": 9879
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.517058678656953e-05,
+      "loss": 15.9172,
+      "step": 9880
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.516739300976959e-05,
+      "loss": 18.3433,
+      "step": 9881
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.516419894898355e-05,
+      "loss": 17.798,
+      "step": 9882
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.51610046042372e-05,
+      "loss": 16.5931,
+      "step": 9883
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.515780997555635e-05,
+      "loss": 16.4746,
+      "step": 9884
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.515461506296682e-05,
+      "loss": 14.7471,
+      "step": 9885
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.515141986649435e-05,
+      "loss": 14.81,
+      "step": 9886
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.51482243861648e-05,
+      "loss": 16.3114,
+      "step": 9887
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.514502862200394e-05,
+      "loss": 16.7175,
+      "step": 9888
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.514183257403759e-05,
+      "loss": 14.389,
+      "step": 9889
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.513863624229154e-05,
+      "loss": 17.0165,
+      "step": 9890
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.513543962679163e-05,
+      "loss": 15.8787,
+      "step": 9891
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.513224272756363e-05,
+      "loss": 14.3519,
+      "step": 9892
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.512904554463342e-05,
+      "loss": 16.8738,
+      "step": 9893
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.512584807802677e-05,
+      "loss": 17.2266,
+      "step": 9894
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.512265032776953e-05,
+      "loss": 16.4333,
+      "step": 9895
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.511945229388749e-05,
+      "loss": 16.3411,
+      "step": 9896
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.511625397640649e-05,
+      "loss": 15.928,
+      "step": 9897
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.511305537535237e-05,
+      "loss": 18.492,
+      "step": 9898
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.510985649075094e-05,
+      "loss": 16.6174,
+      "step": 9899
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.510665732262803e-05,
+      "loss": 15.8597,
+      "step": 9900
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.51034578710095e-05,
+      "loss": 15.0024,
+      "step": 9901
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.510025813592118e-05,
+      "loss": 15.6302,
+      "step": 9902
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.509705811738889e-05,
+      "loss": 14.136,
+      "step": 9903
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.509385781543847e-05,
+      "loss": 16.8154,
+      "step": 9904
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.509065723009577e-05,
+      "loss": 16.9571,
+      "step": 9905
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.508745636138666e-05,
+      "loss": 15.733,
+      "step": 9906
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.508425520933695e-05,
+      "loss": 16.1134,
+      "step": 9907
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.508105377397253e-05,
+      "loss": 16.4863,
+      "step": 9908
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.507785205531922e-05,
+      "loss": 15.5471,
+      "step": 9909
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.507465005340288e-05,
+      "loss": 16.5607,
+      "step": 9910
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.507144776824936e-05,
+      "loss": 16.1325,
+      "step": 9911
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.506824519988455e-05,
+      "loss": 14.7289,
+      "step": 9912
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.506504234833428e-05,
+      "loss": 15.3981,
+      "step": 9913
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.506183921362443e-05,
+      "loss": 16.5063,
+      "step": 9914
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.505863579578088e-05,
+      "loss": 15.9732,
+      "step": 9915
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.505543209482946e-05,
+      "loss": 16.4077,
+      "step": 9916
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.505222811079608e-05,
+      "loss": 16.6582,
+      "step": 9917
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.504902384370657e-05,
+      "loss": 16.9379,
+      "step": 9918
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.504581929358684e-05,
+      "loss": 17.5586,
+      "step": 9919
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.504261446046275e-05,
+      "loss": 15.5547,
+      "step": 9920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.50394093443602e-05,
+      "loss": 14.0092,
+      "step": 9921
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.503620394530507e-05,
+      "loss": 15.8518,
+      "step": 9922
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.503299826332322e-05,
+      "loss": 16.1549,
+      "step": 9923
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.502979229844055e-05,
+      "loss": 16.0892,
+      "step": 9924
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.502658605068297e-05,
+      "loss": 15.4121,
+      "step": 9925
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.502337952007632e-05,
+      "loss": 16.1573,
+      "step": 9926
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.502017270664655e-05,
+      "loss": 18.0705,
+      "step": 9927
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.50169656104195e-05,
+      "loss": 15.9527,
+      "step": 9928
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.501375823142113e-05,
+      "loss": 16.0328,
+      "step": 9929
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.50105505696773e-05,
+      "loss": 15.5211,
+      "step": 9930
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.500734262521392e-05,
+      "loss": 14.6316,
+      "step": 9931
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.50041343980569e-05,
+      "loss": 17.5994,
+      "step": 9932
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.500092588823214e-05,
+      "loss": 16.4452,
+      "step": 9933
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.499771709576557e-05,
+      "loss": 17.6303,
+      "step": 9934
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.499450802068308e-05,
+      "loss": 14.198,
+      "step": 9935
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.499129866301057e-05,
+      "loss": 16.1587,
+      "step": 9936
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.498808902277398e-05,
+      "loss": 16.2023,
+      "step": 9937
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.498487909999924e-05,
+      "loss": 14.7191,
+      "step": 9938
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.498166889471225e-05,
+      "loss": 19.5837,
+      "step": 9939
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.497845840693894e-05,
+      "loss": 16.5502,
+      "step": 9940
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.497524763670524e-05,
+      "loss": 14.8401,
+      "step": 9941
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.497203658403707e-05,
+      "loss": 16.6047,
+      "step": 9942
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.496882524896034e-05,
+      "loss": 17.8043,
+      "step": 9943
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.496561363150104e-05,
+      "loss": 14.8935,
+      "step": 9944
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.496240173168505e-05,
+      "loss": 15.1764,
+      "step": 9945
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.495918954953833e-05,
+      "loss": 15.0193,
+      "step": 9946
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.495597708508682e-05,
+      "loss": 14.4347,
+      "step": 9947
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.495276433835646e-05,
+      "loss": 15.4627,
+      "step": 9948
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.494955130937318e-05,
+      "loss": 15.9095,
+      "step": 9949
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.494633799816296e-05,
+      "loss": 15.6092,
+      "step": 9950
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.49431244047517e-05,
+      "loss": 18.9279,
+      "step": 9951
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.493991052916539e-05,
+      "loss": 16.1718,
+      "step": 9952
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.493669637142996e-05,
+      "loss": 15.9955,
+      "step": 9953
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.493348193157137e-05,
+      "loss": 16.6254,
+      "step": 9954
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.493026720961561e-05,
+      "loss": 16.8755,
+      "step": 9955
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.492705220558859e-05,
+      "loss": 16.7434,
+      "step": 9956
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.49238369195163e-05,
+      "loss": 14.8006,
+      "step": 9957
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.492062135142469e-05,
+      "loss": 15.8565,
+      "step": 9958
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.491740550133975e-05,
+      "loss": 14.6814,
+      "step": 9959
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.491418936928742e-05,
+      "loss": 15.5863,
+      "step": 9960
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.491097295529368e-05,
+      "loss": 17.5916,
+      "step": 9961
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.490775625938452e-05,
+      "loss": 14.6604,
+      "step": 9962
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.49045392815859e-05,
+      "loss": 15.16,
+      "step": 9963
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.490132202192381e-05,
+      "loss": 15.2779,
+      "step": 9964
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.489810448042422e-05,
+      "loss": 16.6596,
+      "step": 9965
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.489488665711309e-05,
+      "loss": 16.2162,
+      "step": 9966
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.489166855201645e-05,
+      "loss": 16.0355,
+      "step": 9967
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.488845016516027e-05,
+      "loss": 16.114,
+      "step": 9968
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.488523149657053e-05,
+      "loss": 16.1005,
+      "step": 9969
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.488201254627323e-05,
+      "loss": 16.2346,
+      "step": 9970
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.487879331429435e-05,
+      "loss": 15.257,
+      "step": 9971
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.487557380065992e-05,
+      "loss": 18.1409,
+      "step": 9972
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.487235400539589e-05,
+      "loss": 16.7506,
+      "step": 9973
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.48691339285283e-05,
+      "loss": 15.6232,
+      "step": 9974
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.486591357008313e-05,
+      "loss": 16.335,
+      "step": 9975
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.48626929300864e-05,
+      "loss": 15.7793,
+      "step": 9976
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.48594720085641e-05,
+      "loss": 14.2007,
+      "step": 9977
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.485625080554227e-05,
+      "loss": 16.7664,
+      "step": 9978
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.48530293210469e-05,
+      "loss": 15.1091,
+      "step": 9979
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.4849807555104e-05,
+      "loss": 16.8343,
+      "step": 9980
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.48465855077396e-05,
+      "loss": 16.4748,
+      "step": 9981
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.484336317897973e-05,
+      "loss": 15.4169,
+      "step": 9982
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.484014056885038e-05,
+      "loss": 14.3295,
+      "step": 9983
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.483691767737758e-05,
+      "loss": 15.0525,
+      "step": 9984
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.483369450458738e-05,
+      "loss": 13.863,
+      "step": 9985
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.483047105050578e-05,
+      "loss": 14.9747,
+      "step": 9986
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.482724731515884e-05,
+      "loss": 15.2215,
+      "step": 9987
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.482402329857256e-05,
+      "loss": 15.243,
+      "step": 9988
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.482079900077298e-05,
+      "loss": 17.0414,
+      "step": 9989
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.481757442178616e-05,
+      "loss": 15.3732,
+      "step": 9990
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.481434956163812e-05,
+      "loss": 16.3148,
+      "step": 9991
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.481112442035492e-05,
+      "loss": 16.1547,
+      "step": 9992
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.480789899796257e-05,
+      "loss": 16.4148,
+      "step": 9993
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.480467329448716e-05,
+      "loss": 15.7822,
+      "step": 9994
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.48014473099547e-05,
+      "loss": 15.2342,
+      "step": 9995
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.479822104439127e-05,
+      "loss": 17.7267,
+      "step": 9996
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.479499449782291e-05,
+      "loss": 17.8354,
+      "step": 9997
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.479176767027566e-05,
+      "loss": 15.2217,
+      "step": 9998
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.478854056177561e-05,
+      "loss": 15.4726,
+      "step": 9999
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.47853131723488e-05,
+      "loss": 16.2776,
+      "step": 10000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.478208550202126e-05,
+      "loss": 15.3616,
+      "step": 10001
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.477885755081914e-05,
+      "loss": 14.1591,
+      "step": 10002
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.477562931876843e-05,
+      "loss": 16.7685,
+      "step": 10003
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.477240080589521e-05,
+      "loss": 15.1013,
+      "step": 10004
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.476917201222559e-05,
+      "loss": 18.2091,
+      "step": 10005
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.476594293778561e-05,
+      "loss": 15.4706,
+      "step": 10006
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.476271358260135e-05,
+      "loss": 16.2355,
+      "step": 10007
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.47594839466989e-05,
+      "loss": 17.2973,
+      "step": 10008
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.475625403010432e-05,
+      "loss": 16.5097,
+      "step": 10009
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.475302383284372e-05,
+      "loss": 14.2975,
+      "step": 10010
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.474979335494315e-05,
+      "loss": 15.2667,
+      "step": 10011
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.474656259642873e-05,
+      "loss": 15.7653,
+      "step": 10012
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.474333155732654e-05,
+      "loss": 17.0198,
+      "step": 10013
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.474010023766266e-05,
+      "loss": 13.8306,
+      "step": 10014
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.47368686374632e-05,
+      "loss": 17.8396,
+      "step": 10015
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.473363675675423e-05,
+      "loss": 16.4185,
+      "step": 10016
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.473040459556187e-05,
+      "loss": 17.5181,
+      "step": 10017
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.472717215391221e-05,
+      "loss": 14.1967,
+      "step": 10018
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.472393943183138e-05,
+      "loss": 14.567,
+      "step": 10019
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.472070642934546e-05,
+      "loss": 16.7171,
+      "step": 10020
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.471747314648054e-05,
+      "loss": 16.8782,
+      "step": 10021
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.471423958326276e-05,
+      "loss": 16.8097,
+      "step": 10022
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.471100573971823e-05,
+      "loss": 14.2488,
+      "step": 10023
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.470777161587304e-05,
+      "loss": 16.8012,
+      "step": 10024
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.470453721175335e-05,
+      "loss": 16.0408,
+      "step": 10025
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.470130252738524e-05,
+      "loss": 15.3317,
+      "step": 10026
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.469806756279483e-05,
+      "loss": 16.3397,
+      "step": 10027
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.469483231800827e-05,
+      "loss": 16.6813,
+      "step": 10028
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.469159679305167e-05,
+      "loss": 15.3972,
+      "step": 10029
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.468836098795115e-05,
+      "loss": 17.7924,
+      "step": 10030
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.468512490273286e-05,
+      "loss": 16.1695,
+      "step": 10031
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.468188853742292e-05,
+      "loss": 15.5341,
+      "step": 10032
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.467865189204744e-05,
+      "loss": 15.4826,
+      "step": 10033
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.467541496663261e-05,
+      "loss": 16.5606,
+      "step": 10034
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.467217776120454e-05,
+      "loss": 16.6002,
+      "step": 10035
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.466894027578937e-05,
+      "loss": 16.8228,
+      "step": 10036
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.466570251041324e-05,
+      "loss": 18.3065,
+      "step": 10037
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.466246446510231e-05,
+      "loss": 15.198,
+      "step": 10038
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.465922613988272e-05,
+      "loss": 18.3757,
+      "step": 10039
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.465598753478062e-05,
+      "loss": 14.0999,
+      "step": 10040
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.465274864982215e-05,
+      "loss": 16.7963,
+      "step": 10041
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.464950948503349e-05,
+      "loss": 16.1525,
+      "step": 10042
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.464627004044079e-05,
+      "loss": 15.8685,
+      "step": 10043
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.46430303160702e-05,
+      "loss": 16.7289,
+      "step": 10044
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.463979031194788e-05,
+      "loss": 16.1837,
+      "step": 10045
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.463655002810003e-05,
+      "loss": 15.5,
+      "step": 10046
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.463330946455274e-05,
+      "loss": 14.7962,
+      "step": 10047
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.463006862133226e-05,
+      "loss": 14.9811,
+      "step": 10048
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.462682749846471e-05,
+      "loss": 15.7021,
+      "step": 10049
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.46235860959763e-05,
+      "loss": 17.6185,
+      "step": 10050
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.462034441389314e-05,
+      "loss": 17.5102,
+      "step": 10051
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.461710245224148e-05,
+      "loss": 15.7741,
+      "step": 10052
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.461386021104746e-05,
+      "loss": 15.4966,
+      "step": 10053
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.461061769033729e-05,
+      "loss": 16.9158,
+      "step": 10054
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.460737489013713e-05,
+      "loss": 15.4015,
+      "step": 10055
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.460413181047317e-05,
+      "loss": 14.246,
+      "step": 10056
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.46008884513716e-05,
+      "loss": 15.1627,
+      "step": 10057
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.459764481285863e-05,
+      "loss": 16.1755,
+      "step": 10058
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.459440089496043e-05,
+      "loss": 15.1312,
+      "step": 10059
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.459115669770318e-05,
+      "loss": 15.8356,
+      "step": 10060
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.458791222111312e-05,
+      "loss": 15.2699,
+      "step": 10061
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.458466746521643e-05,
+      "loss": 18.0383,
+      "step": 10062
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.458142243003931e-05,
+      "loss": 16.1557,
+      "step": 10063
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.457817711560796e-05,
+      "loss": 14.3668,
+      "step": 10064
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.457493152194859e-05,
+      "loss": 15.8139,
+      "step": 10065
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.457168564908742e-05,
+      "loss": 14.3766,
+      "step": 10066
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.456843949705064e-05,
+      "loss": 15.5619,
+      "step": 10067
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.45651930658645e-05,
+      "loss": 16.4941,
+      "step": 10068
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.456194635555518e-05,
+      "loss": 17.1351,
+      "step": 10069
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.455869936614891e-05,
+      "loss": 14.6663,
+      "step": 10070
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.45554520976719e-05,
+      "loss": 15.9028,
+      "step": 10071
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.45522045501504e-05,
+      "loss": 15.9512,
+      "step": 10072
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.454895672361063e-05,
+      "loss": 15.6606,
+      "step": 10073
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.454570861807879e-05,
+      "loss": 16.5766,
+      "step": 10074
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.454246023358112e-05,
+      "loss": 18.3728,
+      "step": 10075
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.453921157014388e-05,
+      "loss": 15.8388,
+      "step": 10076
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.453596262779326e-05,
+      "loss": 16.1321,
+      "step": 10077
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.45327134065555e-05,
+      "loss": 16.9528,
+      "step": 10078
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.452946390645689e-05,
+      "loss": 16.1547,
+      "step": 10079
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.452621412752361e-05,
+      "loss": 14.639,
+      "step": 10080
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.452296406978195e-05,
+      "loss": 15.6613,
+      "step": 10081
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.451971373325812e-05,
+      "loss": 15.187,
+      "step": 10082
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.451646311797839e-05,
+      "loss": 16.2391,
+      "step": 10083
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.4513212223969e-05,
+      "loss": 16.5556,
+      "step": 10084
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.45099610512562e-05,
+      "loss": 16.3477,
+      "step": 10085
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.450670959986622e-05,
+      "loss": 15.9711,
+      "step": 10086
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.450345786982536e-05,
+      "loss": 15.6085,
+      "step": 10087
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.450020586115987e-05,
+      "loss": 16.4547,
+      "step": 10088
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.449695357389599e-05,
+      "loss": 16.2535,
+      "step": 10089
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.449370100806002e-05,
+      "loss": 14.9911,
+      "step": 10090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.449044816367818e-05,
+      "loss": 15.7001,
+      "step": 10091
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.448719504077676e-05,
+      "loss": 14.0718,
+      "step": 10092
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.448394163938203e-05,
+      "loss": 15.5069,
+      "step": 10093
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.448068795952026e-05,
+      "loss": 16.0816,
+      "step": 10094
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.447743400121772e-05,
+      "loss": 17.1373,
+      "step": 10095
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.44741797645007e-05,
+      "loss": 17.5118,
+      "step": 10096
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.447092524939546e-05,
+      "loss": 15.5789,
+      "step": 10097
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.44676704559283e-05,
+      "loss": 15.181,
+      "step": 10098
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.446441538412549e-05,
+      "loss": 15.0699,
+      "step": 10099
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.446116003401332e-05,
+      "loss": 14.5795,
+      "step": 10100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.445790440561807e-05,
+      "loss": 15.3287,
+      "step": 10101
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.445464849896605e-05,
+      "loss": 15.8972,
+      "step": 10102
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.445139231408352e-05,
+      "loss": 14.5589,
+      "step": 10103
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.444813585099682e-05,
+      "loss": 15.1813,
+      "step": 10104
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.44448791097322e-05,
+      "loss": 15.1807,
+      "step": 10105
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.444162209031599e-05,
+      "loss": 13.6691,
+      "step": 10106
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.443836479277449e-05,
+      "loss": 15.2601,
+      "step": 10107
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.443510721713398e-05,
+      "loss": 19.0413,
+      "step": 10108
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.443184936342079e-05,
+      "loss": 14.9692,
+      "step": 10109
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.442859123166122e-05,
+      "loss": 14.3651,
+      "step": 10110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.442533282188156e-05,
+      "loss": 14.3072,
+      "step": 10111
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.442207413410817e-05,
+      "loss": 16.4398,
+      "step": 10112
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.441881516836731e-05,
+      "loss": 14.5623,
+      "step": 10113
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.441555592468534e-05,
+      "loss": 16.4615,
+      "step": 10114
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.441229640308856e-05,
+      "loss": 16.6283,
+      "step": 10115
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.44090366036033e-05,
+      "loss": 16.4433,
+      "step": 10116
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.440577652625587e-05,
+      "loss": 15.4385,
+      "step": 10117
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.44025161710726e-05,
+      "loss": 14.5858,
+      "step": 10118
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.439925553807983e-05,
+      "loss": 15.7016,
+      "step": 10119
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.439599462730388e-05,
+      "loss": 15.6253,
+      "step": 10120
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.439273343877109e-05,
+      "loss": 15.8749,
+      "step": 10121
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.43894719725078e-05,
+      "loss": 14.8057,
+      "step": 10122
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.438621022854032e-05,
+      "loss": 15.2226,
+      "step": 10123
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.4382948206895e-05,
+      "loss": 15.1053,
+      "step": 10124
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.43796859075982e-05,
+      "loss": 16.5915,
+      "step": 10125
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.437642333067625e-05,
+      "loss": 16.347,
+      "step": 10126
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.437316047615552e-05,
+      "loss": 15.5389,
+      "step": 10127
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.436989734406232e-05,
+      "loss": 15.669,
+      "step": 10128
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.4366633934423e-05,
+      "loss": 16.0551,
+      "step": 10129
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.436337024726395e-05,
+      "loss": 16.3394,
+      "step": 10130
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.436010628261151e-05,
+      "loss": 15.7925,
+      "step": 10131
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.435684204049204e-05,
+      "loss": 17.1697,
+      "step": 10132
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.435357752093187e-05,
+      "loss": 15.9083,
+      "step": 10133
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.43503127239574e-05,
+      "loss": 14.9348,
+      "step": 10134
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.434704764959497e-05,
+      "loss": 17.4945,
+      "step": 10135
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.434378229787096e-05,
+      "loss": 16.795,
+      "step": 10136
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.434051666881173e-05,
+      "loss": 14.5373,
+      "step": 10137
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.433725076244368e-05,
+      "loss": 14.2802,
+      "step": 10138
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.433398457879312e-05,
+      "loss": 15.8029,
+      "step": 10139
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.433071811788649e-05,
+      "loss": 16.9602,
+      "step": 10140
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.432745137975013e-05,
+      "loss": 15.5683,
+      "step": 10141
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.432418436441043e-05,
+      "loss": 14.5639,
+      "step": 10142
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.432091707189379e-05,
+      "loss": 15.1849,
+      "step": 10143
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.431764950222656e-05,
+      "loss": 16.0714,
+      "step": 10144
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.431438165543514e-05,
+      "loss": 16.19,
+      "step": 10145
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.431111353154593e-05,
+      "loss": 16.1786,
+      "step": 10146
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.430784513058531e-05,
+      "loss": 15.7554,
+      "step": 10147
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.430457645257968e-05,
+      "loss": 16.4254,
+      "step": 10148
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.430130749755543e-05,
+      "loss": 17.8419,
+      "step": 10149
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.429803826553894e-05,
+      "loss": 14.8702,
+      "step": 10150
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.429476875655667e-05,
+      "loss": 16.2014,
+      "step": 10151
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.429149897063495e-05,
+      "loss": 16.4565,
+      "step": 10152
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.428822890780022e-05,
+      "loss": 16.414,
+      "step": 10153
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.428495856807888e-05,
+      "loss": 15.611,
+      "step": 10154
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.428168795149735e-05,
+      "loss": 16.5748,
+      "step": 10155
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.427841705808203e-05,
+      "loss": 17.7021,
+      "step": 10156
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.427514588785934e-05,
+      "loss": 16.2182,
+      "step": 10157
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.42718744408557e-05,
+      "loss": 15.2637,
+      "step": 10158
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.42686027170975e-05,
+      "loss": 14.8826,
+      "step": 10159
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.426533071661118e-05,
+      "loss": 15.8045,
+      "step": 10160
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.426205843942317e-05,
+      "loss": 16.8872,
+      "step": 10161
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.425878588555988e-05,
+      "loss": 17.0128,
+      "step": 10162
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.425551305504776e-05,
+      "loss": 16.6174,
+      "step": 10163
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.42522399479132e-05,
+      "loss": 17.4531,
+      "step": 10164
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.424896656418267e-05,
+      "loss": 15.0596,
+      "step": 10165
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.424569290388258e-05,
+      "loss": 15.2452,
+      "step": 10166
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.424241896703936e-05,
+      "loss": 16.2991,
+      "step": 10167
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.423914475367947e-05,
+      "loss": 16.3533,
+      "step": 10168
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.423587026382934e-05,
+      "loss": 16.4514,
+      "step": 10169
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.423259549751541e-05,
+      "loss": 17.9653,
+      "step": 10170
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.422932045476413e-05,
+      "loss": 16.2255,
+      "step": 10171
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.422604513560195e-05,
+      "loss": 13.8814,
+      "step": 10172
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.42227695400553e-05,
+      "loss": 16.9911,
+      "step": 10173
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.421949366815064e-05,
+      "loss": 16.4865,
+      "step": 10174
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.421621751991446e-05,
+      "loss": 17.0993,
+      "step": 10175
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.421294109537315e-05,
+      "loss": 14.1109,
+      "step": 10176
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.420966439455323e-05,
+      "loss": 14.4305,
+      "step": 10177
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.42063874174811e-05,
+      "loss": 16.4271,
+      "step": 10178
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.420311016418327e-05,
+      "loss": 17.031,
+      "step": 10179
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.41998326346862e-05,
+      "loss": 16.2603,
+      "step": 10180
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.419655482901634e-05,
+      "loss": 15.7146,
+      "step": 10181
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.419327674720016e-05,
+      "loss": 15.5275,
+      "step": 10182
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.418999838926414e-05,
+      "loss": 15.7496,
+      "step": 10183
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.418671975523474e-05,
+      "loss": 17.4148,
+      "step": 10184
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.418344084513845e-05,
+      "loss": 16.3339,
+      "step": 10185
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.418016165900176e-05,
+      "loss": 17.5145,
+      "step": 10186
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.417688219685112e-05,
+      "loss": 16.8972,
+      "step": 10187
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.417360245871303e-05,
+      "loss": 15.7481,
+      "step": 10188
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.417032244461398e-05,
+      "loss": 16.5057,
+      "step": 10189
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.416704215458043e-05,
+      "loss": 15.2582,
+      "step": 10190
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.41637615886389e-05,
+      "loss": 16.3508,
+      "step": 10191
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.416048074681586e-05,
+      "loss": 17.6886,
+      "step": 10192
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.415719962913782e-05,
+      "loss": 15.044,
+      "step": 10193
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.415391823563128e-05,
+      "loss": 15.4058,
+      "step": 10194
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.415063656632271e-05,
+      "loss": 16.4548,
+      "step": 10195
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.414735462123864e-05,
+      "loss": 16.2206,
+      "step": 10196
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.414407240040555e-05,
+      "loss": 15.9985,
+      "step": 10197
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.414078990384995e-05,
+      "loss": 16.9887,
+      "step": 10198
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.413750713159836e-05,
+      "loss": 15.7197,
+      "step": 10199
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.413422408367728e-05,
+      "loss": 17.8878,
+      "step": 10200
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.413094076011323e-05,
+      "loss": 16.9812,
+      "step": 10201
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.412765716093272e-05,
+      "loss": 17.0704,
+      "step": 10202
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.412437328616224e-05,
+      "loss": 15.7244,
+      "step": 10203
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.412108913582835e-05,
+      "loss": 15.6955,
+      "step": 10204
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.411780470995753e-05,
+      "loss": 17.3508,
+      "step": 10205
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.411452000857635e-05,
+      "loss": 17.4853,
+      "step": 10206
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.411123503171129e-05,
+      "loss": 16.9892,
+      "step": 10207
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.41079497793889e-05,
+      "loss": 15.2212,
+      "step": 10208
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.41046642516357e-05,
+      "loss": 15.2525,
+      "step": 10209
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.410137844847823e-05,
+      "loss": 15.9828,
+      "step": 10210
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.409809236994303e-05,
+      "loss": 15.4875,
+      "step": 10211
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.409480601605662e-05,
+      "loss": 16.0024,
+      "step": 10212
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.409151938684554e-05,
+      "loss": 15.6829,
+      "step": 10213
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.408823248233633e-05,
+      "loss": 15.7301,
+      "step": 10214
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.408494530255556e-05,
+      "loss": 15.3036,
+      "step": 10215
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.408165784752973e-05,
+      "loss": 17.1446,
+      "step": 10216
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.407837011728542e-05,
+      "loss": 16.0392,
+      "step": 10217
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.407508211184915e-05,
+      "loss": 15.2463,
+      "step": 10218
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.40717938312475e-05,
+      "loss": 17.5023,
+      "step": 10219
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.4068505275507e-05,
+      "loss": 16.1231,
+      "step": 10220
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.406521644465423e-05,
+      "loss": 15.2347,
+      "step": 10221
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.406192733871575e-05,
+      "loss": 14.974,
+      "step": 10222
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.40586379577181e-05,
+      "loss": 15.2186,
+      "step": 10223
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.405534830168785e-05,
+      "loss": 16.1281,
+      "step": 10224
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.405205837065157e-05,
+      "loss": 15.561,
+      "step": 10225
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.404876816463583e-05,
+      "loss": 15.4801,
+      "step": 10226
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.404547768366717e-05,
+      "loss": 17.1948,
+      "step": 10227
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.404218692777219e-05,
+      "loss": 15.2767,
+      "step": 10228
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.403889589697746e-05,
+      "loss": 16.6099,
+      "step": 10229
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.403560459130957e-05,
+      "loss": 16.0596,
+      "step": 10230
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.403231301079507e-05,
+      "loss": 14.8848,
+      "step": 10231
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.402902115546054e-05,
+      "loss": 16.1324,
+      "step": 10232
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.402572902533258e-05,
+      "loss": 16.0611,
+      "step": 10233
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.402243662043778e-05,
+      "loss": 15.9239,
+      "step": 10234
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.401914394080271e-05,
+      "loss": 17.2604,
+      "step": 10235
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.401585098645395e-05,
+      "loss": 16.4192,
+      "step": 10236
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.401255775741813e-05,
+      "loss": 15.0109,
+      "step": 10237
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.400926425372182e-05,
+      "loss": 16.9966,
+      "step": 10238
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.400597047539161e-05,
+      "loss": 17.3327,
+      "step": 10239
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.40026764224541e-05,
+      "loss": 17.0052,
+      "step": 10240
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.39993820949359e-05,
+      "loss": 15.9959,
+      "step": 10241
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.39960874928636e-05,
+      "loss": 16.892,
+      "step": 10242
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.399279261626382e-05,
+      "loss": 17.3706,
+      "step": 10243
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.398949746516316e-05,
+      "loss": 17.4239,
+      "step": 10244
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.398620203958823e-05,
+      "loss": 15.0952,
+      "step": 10245
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.398290633956562e-05,
+      "loss": 18.2461,
+      "step": 10246
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.397961036512199e-05,
+      "loss": 16.0332,
+      "step": 10247
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.397631411628391e-05,
+      "loss": 15.0798,
+      "step": 10248
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.397301759307803e-05,
+      "loss": 15.3108,
+      "step": 10249
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.396972079553096e-05,
+      "loss": 16.2252,
+      "step": 10250
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.396642372366931e-05,
+      "loss": 17.5808,
+      "step": 10251
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.396312637751973e-05,
+      "loss": 16.6659,
+      "step": 10252
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.395982875710881e-05,
+      "loss": 15.4672,
+      "step": 10253
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.395653086246322e-05,
+      "loss": 17.2025,
+      "step": 10254
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.395323269360958e-05,
+      "loss": 17.829,
+      "step": 10255
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.394993425057449e-05,
+      "loss": 15.7299,
+      "step": 10256
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.394663553338463e-05,
+      "loss": 17.007,
+      "step": 10257
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.394333654206661e-05,
+      "loss": 15.1415,
+      "step": 10258
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.39400372766471e-05,
+      "loss": 17.4645,
+      "step": 10259
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.39367377371527e-05,
+      "loss": 17.459,
+      "step": 10260
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.393343792361009e-05,
+      "loss": 17.1014,
+      "step": 10261
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.393013783604591e-05,
+      "loss": 16.0013,
+      "step": 10262
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.392683747448678e-05,
+      "loss": 16.6128,
+      "step": 10263
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.39235368389594e-05,
+      "loss": 16.947,
+      "step": 10264
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.392023592949039e-05,
+      "loss": 15.4529,
+      "step": 10265
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.39169347461064e-05,
+      "loss": 16.0279,
+      "step": 10266
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.391363328883413e-05,
+      "loss": 16.3946,
+      "step": 10267
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.391033155770019e-05,
+      "loss": 16.3241,
+      "step": 10268
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.390702955273127e-05,
+      "loss": 15.9315,
+      "step": 10269
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.390372727395403e-05,
+      "loss": 14.8237,
+      "step": 10270
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.390042472139515e-05,
+      "loss": 18.1196,
+      "step": 10271
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.389712189508127e-05,
+      "loss": 17.0454,
+      "step": 10272
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.389381879503909e-05,
+      "loss": 15.7753,
+      "step": 10273
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.389051542129524e-05,
+      "loss": 16.3553,
+      "step": 10274
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.388721177387643e-05,
+      "loss": 16.9119,
+      "step": 10275
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.388390785280936e-05,
+      "loss": 15.8957,
+      "step": 10276
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.388060365812067e-05,
+      "loss": 17.3901,
+      "step": 10277
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.387729918983706e-05,
+      "loss": 14.7762,
+      "step": 10278
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.38739944479852e-05,
+      "loss": 15.3449,
+      "step": 10279
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.387068943259178e-05,
+      "loss": 14.9715,
+      "step": 10280
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.386738414368351e-05,
+      "loss": 17.4647,
+      "step": 10281
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.386407858128706e-05,
+      "loss": 17.6097,
+      "step": 10282
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.386077274542913e-05,
+      "loss": 15.9654,
+      "step": 10283
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.385746663613642e-05,
+      "loss": 15.29,
+      "step": 10284
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.385416025343562e-05,
+      "loss": 15.9319,
+      "step": 10285
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.385085359735342e-05,
+      "loss": 15.7385,
+      "step": 10286
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.384754666791655e-05,
+      "loss": 16.1795,
+      "step": 10287
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.384423946515169e-05,
+      "loss": 16.3886,
+      "step": 10288
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.384093198908558e-05,
+      "loss": 16.1736,
+      "step": 10289
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.383762423974489e-05,
+      "loss": 14.5426,
+      "step": 10290
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.383431621715634e-05,
+      "loss": 17.0959,
+      "step": 10291
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.383100792134663e-05,
+      "loss": 16.7186,
+      "step": 10292
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.382769935234254e-05,
+      "loss": 15.5607,
+      "step": 10293
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.38243905101707e-05,
+      "loss": 17.0271,
+      "step": 10294
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.382108139485789e-05,
+      "loss": 14.2281,
+      "step": 10295
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.381777200643082e-05,
+      "loss": 15.6288,
+      "step": 10296
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.381446234491618e-05,
+      "loss": 15.2269,
+      "step": 10297
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.381115241034076e-05,
+      "loss": 16.9246,
+      "step": 10298
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.380784220273124e-05,
+      "loss": 16.0199,
+      "step": 10299
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.380453172211435e-05,
+      "loss": 15.4932,
+      "step": 10300
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.380122096851686e-05,
+      "loss": 15.1219,
+      "step": 10301
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.379790994196546e-05,
+      "loss": 16.8071,
+      "step": 10302
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.37945986424869e-05,
+      "loss": 15.1976,
+      "step": 10303
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.379128707010796e-05,
+      "loss": 13.9399,
+      "step": 10304
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.378797522485533e-05,
+      "loss": 15.7751,
+      "step": 10305
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.37846631067558e-05,
+      "loss": 15.4722,
+      "step": 10306
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.378135071583606e-05,
+      "loss": 15.2636,
+      "step": 10307
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.37780380521229e-05,
+      "loss": 14.8531,
+      "step": 10308
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.377472511564305e-05,
+      "loss": 16.6009,
+      "step": 10309
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.37714119064233e-05,
+      "loss": 14.5639,
+      "step": 10310
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.376809842449036e-05,
+      "loss": 16.7203,
+      "step": 10311
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.3764784669871e-05,
+      "loss": 14.7066,
+      "step": 10312
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.376147064259198e-05,
+      "loss": 15.7836,
+      "step": 10313
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.375815634268009e-05,
+      "loss": 16.2974,
+      "step": 10314
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.375484177016206e-05,
+      "loss": 14.6204,
+      "step": 10315
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.375152692506468e-05,
+      "loss": 15.4978,
+      "step": 10316
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.374821180741468e-05,
+      "loss": 17.0517,
+      "step": 10317
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.374489641723886e-05,
+      "loss": 16.806,
+      "step": 10318
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.3741580754564e-05,
+      "loss": 18.0535,
+      "step": 10319
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.373826481941685e-05,
+      "loss": 15.2685,
+      "step": 10320
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.373494861182419e-05,
+      "loss": 16.3535,
+      "step": 10321
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.373163213181283e-05,
+      "loss": 16.379,
+      "step": 10322
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.372831537940951e-05,
+      "loss": 15.7547,
+      "step": 10323
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.372499835464105e-05,
+      "loss": 14.8239,
+      "step": 10324
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.372168105753421e-05,
+      "loss": 14.1058,
+      "step": 10325
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.371836348811579e-05,
+      "loss": 15.9763,
+      "step": 10326
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.371504564641259e-05,
+      "loss": 15.859,
+      "step": 10327
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.371172753245138e-05,
+      "loss": 17.6639,
+      "step": 10328
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.370840914625897e-05,
+      "loss": 14.6686,
+      "step": 10329
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.370509048786214e-05,
+      "loss": 14.0107,
+      "step": 10330
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.370177155728771e-05,
+      "loss": 15.8753,
+      "step": 10331
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.369845235456246e-05,
+      "loss": 14.2397,
+      "step": 10332
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.369513287971322e-05,
+      "loss": 16.3509,
+      "step": 10333
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.369181313276679e-05,
+      "loss": 15.9349,
+      "step": 10334
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.368849311374995e-05,
+      "loss": 15.4912,
+      "step": 10335
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.368517282268953e-05,
+      "loss": 17.0995,
+      "step": 10336
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.368185225961236e-05,
+      "loss": 14.6419,
+      "step": 10337
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.367853142454522e-05,
+      "loss": 17.5445,
+      "step": 10338
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.367521031751494e-05,
+      "loss": 15.6235,
+      "step": 10339
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.367188893854835e-05,
+      "loss": 15.2189,
+      "step": 10340
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.366856728767227e-05,
+      "loss": 16.0806,
+      "step": 10341
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.366524536491349e-05,
+      "loss": 16.0399,
+      "step": 10342
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.366192317029888e-05,
+      "loss": 16.1672,
+      "step": 10343
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.365860070385525e-05,
+      "loss": 14.7456,
+      "step": 10344
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.365527796560942e-05,
+      "loss": 15.4918,
+      "step": 10345
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.365195495558824e-05,
+      "loss": 15.9251,
+      "step": 10346
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.364863167381853e-05,
+      "loss": 16.3399,
+      "step": 10347
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.364530812032712e-05,
+      "loss": 15.4127,
+      "step": 10348
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.364198429514087e-05,
+      "loss": 14.6107,
+      "step": 10349
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.363866019828661e-05,
+      "loss": 17.3096,
+      "step": 10350
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.363533582979119e-05,
+      "loss": 17.8155,
+      "step": 10351
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.363201118968144e-05,
+      "loss": 16.213,
+      "step": 10352
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.362868627798421e-05,
+      "loss": 15.5246,
+      "step": 10353
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.362536109472636e-05,
+      "loss": 16.9884,
+      "step": 10354
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.362203563993475e-05,
+      "loss": 15.3099,
+      "step": 10355
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.36187099136362e-05,
+      "loss": 15.8699,
+      "step": 10356
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.36153839158576e-05,
+      "loss": 17.5505,
+      "step": 10357
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.36120576466258e-05,
+      "loss": 15.0823,
+      "step": 10358
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.360873110596765e-05,
+      "loss": 13.7266,
+      "step": 10359
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.360540429391002e-05,
+      "loss": 15.277,
+      "step": 10360
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.360207721047978e-05,
+      "loss": 16.5911,
+      "step": 10361
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.359874985570377e-05,
+      "loss": 17.1973,
+      "step": 10362
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.35954222296089e-05,
+      "loss": 17.4833,
+      "step": 10363
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.359209433222201e-05,
+      "loss": 17.5368,
+      "step": 10364
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.358876616356999e-05,
+      "loss": 17.7728,
+      "step": 10365
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.358543772367971e-05,
+      "loss": 16.4426,
+      "step": 10366
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.358210901257806e-05,
+      "loss": 15.9565,
+      "step": 10367
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.35787800302919e-05,
+      "loss": 15.4813,
+      "step": 10368
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.357545077684813e-05,
+      "loss": 15.693,
+      "step": 10369
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.357212125227361e-05,
+      "loss": 15.5767,
+      "step": 10370
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.356879145659525e-05,
+      "loss": 15.1178,
+      "step": 10371
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.356546138983993e-05,
+      "loss": 16.6829,
+      "step": 10372
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.356213105203454e-05,
+      "loss": 16.3712,
+      "step": 10373
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.355880044320598e-05,
+      "loss": 14.5445,
+      "step": 10374
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.355546956338113e-05,
+      "loss": 17.1845,
+      "step": 10375
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.355213841258694e-05,
+      "loss": 17.4835,
+      "step": 10376
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.354880699085024e-05,
+      "loss": 13.8128,
+      "step": 10377
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.354547529819796e-05,
+      "loss": 16.5529,
+      "step": 10378
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.354214333465702e-05,
+      "loss": 15.2261,
+      "step": 10379
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.35388111002543e-05,
+      "loss": 14.775,
+      "step": 10380
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.353547859501673e-05,
+      "loss": 17.5817,
+      "step": 10381
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.35321458189712e-05,
+      "loss": 15.9243,
+      "step": 10382
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.352881277214465e-05,
+      "loss": 16.7867,
+      "step": 10383
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.352547945456397e-05,
+      "loss": 15.9092,
+      "step": 10384
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.352214586625609e-05,
+      "loss": 18.5354,
+      "step": 10385
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.351881200724794e-05,
+      "loss": 15.5067,
+      "step": 10386
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.351547787756642e-05,
+      "loss": 16.2066,
+      "step": 10387
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.351214347723845e-05,
+      "loss": 15.3674,
+      "step": 10388
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.3508808806291e-05,
+      "loss": 17.5123,
+      "step": 10389
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.350547386475094e-05,
+      "loss": 17.0114,
+      "step": 10390
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.350213865264525e-05,
+      "loss": 15.5072,
+      "step": 10391
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.349880317000082e-05,
+      "loss": 15.7838,
+      "step": 10392
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.349546741684462e-05,
+      "loss": 14.0626,
+      "step": 10393
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.349213139320357e-05,
+      "loss": 14.3427,
+      "step": 10394
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.348879509910462e-05,
+      "loss": 14.6403,
+      "step": 10395
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.34854585345747e-05,
+      "loss": 15.4939,
+      "step": 10396
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.348212169964076e-05,
+      "loss": 16.6833,
+      "step": 10397
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.347878459432973e-05,
+      "loss": 16.0644,
+      "step": 10398
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.347544721866857e-05,
+      "loss": 15.8768,
+      "step": 10399
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.347210957268425e-05,
+      "loss": 14.888,
+      "step": 10400
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.34687716564037e-05,
+      "loss": 16.6275,
+      "step": 10401
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.346543346985387e-05,
+      "loss": 14.8954,
+      "step": 10402
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.346209501306173e-05,
+      "loss": 15.1738,
+      "step": 10403
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.345875628605424e-05,
+      "loss": 15.33,
+      "step": 10404
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.345541728885835e-05,
+      "loss": 14.4809,
+      "step": 10405
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.345207802150103e-05,
+      "loss": 14.3283,
+      "step": 10406
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.344873848400925e-05,
+      "loss": 15.8058,
+      "step": 10407
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.344539867640997e-05,
+      "loss": 16.5901,
+      "step": 10408
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.344205859873015e-05,
+      "loss": 13.6447,
+      "step": 10409
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.343871825099679e-05,
+      "loss": 14.9815,
+      "step": 10410
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.343537763323683e-05,
+      "loss": 14.3955,
+      "step": 10411
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.343203674547728e-05,
+      "loss": 17.5492,
+      "step": 10412
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.342869558774509e-05,
+      "loss": 15.396,
+      "step": 10413
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.342535416006725e-05,
+      "loss": 15.8689,
+      "step": 10414
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.342201246247077e-05,
+      "loss": 17.4723,
+      "step": 10415
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.34186704949826e-05,
+      "loss": 16.262,
+      "step": 10416
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.341532825762972e-05,
+      "loss": 15.6296,
+      "step": 10417
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.341198575043915e-05,
+      "loss": 17.3561,
+      "step": 10418
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.340864297343789e-05,
+      "loss": 16.3598,
+      "step": 10419
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.34052999266529e-05,
+      "loss": 16.9822,
+      "step": 10420
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.340195661011119e-05,
+      "loss": 14.573,
+      "step": 10421
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.339861302383976e-05,
+      "loss": 17.1068,
+      "step": 10422
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.339526916786559e-05,
+      "loss": 15.4079,
+      "step": 10423
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.339192504221571e-05,
+      "loss": 17.313,
+      "step": 10424
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.338858064691713e-05,
+      "loss": 15.6866,
+      "step": 10425
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.338523598199683e-05,
+      "loss": 14.8605,
+      "step": 10426
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.338189104748185e-05,
+      "loss": 15.1745,
+      "step": 10427
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.337854584339917e-05,
+      "loss": 16.314,
+      "step": 10428
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.33752003697758e-05,
+      "loss": 17.2803,
+      "step": 10429
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.337185462663878e-05,
+      "loss": 13.7008,
+      "step": 10430
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.336850861401514e-05,
+      "loss": 16.6301,
+      "step": 10431
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.336516233193187e-05,
+      "loss": 17.7315,
+      "step": 10432
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.3361815780416e-05,
+      "loss": 16.4179,
+      "step": 10433
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.335846895949455e-05,
+      "loss": 17.6159,
+      "step": 10434
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.335512186919455e-05,
+      "loss": 15.7494,
+      "step": 10435
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.335177450954303e-05,
+      "loss": 16.1823,
+      "step": 10436
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.334842688056704e-05,
+      "loss": 15.5946,
+      "step": 10437
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.334507898229357e-05,
+      "loss": 18.1387,
+      "step": 10438
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.334173081474971e-05,
+      "loss": 16.6665,
+      "step": 10439
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.333838237796246e-05,
+      "loss": 16.8682,
+      "step": 10440
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.333503367195884e-05,
+      "loss": 14.6292,
+      "step": 10441
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.333168469676595e-05,
+      "loss": 16.7218,
+      "step": 10442
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.332833545241079e-05,
+      "loss": 14.9269,
+      "step": 10443
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.332498593892042e-05,
+      "loss": 16.6745,
+      "step": 10444
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.332163615632189e-05,
+      "loss": 17.1558,
+      "step": 10445
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.331828610464223e-05,
+      "loss": 14.5907,
+      "step": 10446
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.331493578390851e-05,
+      "loss": 15.6483,
+      "step": 10447
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.331158519414781e-05,
+      "loss": 15.195,
+      "step": 10448
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.330823433538715e-05,
+      "loss": 15.9357,
+      "step": 10449
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.330488320765359e-05,
+      "loss": 16.1036,
+      "step": 10450
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.330153181097423e-05,
+      "loss": 15.4484,
+      "step": 10451
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.329818014537608e-05,
+      "loss": 16.3573,
+      "step": 10452
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.329482821088623e-05,
+      "loss": 16.7993,
+      "step": 10453
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.329147600753176e-05,
+      "loss": 15.5623,
+      "step": 10454
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.328812353533972e-05,
+      "loss": 16.4402,
+      "step": 10455
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.328477079433718e-05,
+      "loss": 17.2954,
+      "step": 10456
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.328141778455125e-05,
+      "loss": 15.522,
+      "step": 10457
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.327806450600897e-05,
+      "loss": 16.6238,
+      "step": 10458
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.327471095873743e-05,
+      "loss": 15.7878,
+      "step": 10459
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.327135714276371e-05,
+      "loss": 17.1973,
+      "step": 10460
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.326800305811488e-05,
+      "loss": 14.9162,
+      "step": 10461
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.326464870481807e-05,
+      "loss": 15.5759,
+      "step": 10462
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.326129408290031e-05,
+      "loss": 17.4654,
+      "step": 10463
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.325793919238873e-05,
+      "loss": 16.7152,
+      "step": 10464
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.325458403331039e-05,
+      "loss": 16.0362,
+      "step": 10465
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.32512286056924e-05,
+      "loss": 16.6722,
+      "step": 10466
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.324787290956189e-05,
+      "loss": 14.2087,
+      "step": 10467
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.32445169449459e-05,
+      "loss": 14.9107,
+      "step": 10468
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.324116071187156e-05,
+      "loss": 14.0608,
+      "step": 10469
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.323780421036596e-05,
+      "loss": 14.4446,
+      "step": 10470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.323444744045624e-05,
+      "loss": 18.1922,
+      "step": 10471
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.323109040216944e-05,
+      "loss": 15.6392,
+      "step": 10472
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.322773309553273e-05,
+      "loss": 16.5603,
+      "step": 10473
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.32243755205732e-05,
+      "loss": 15.9002,
+      "step": 10474
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.322101767731796e-05,
+      "loss": 16.1672,
+      "step": 10475
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.321765956579412e-05,
+      "loss": 15.7673,
+      "step": 10476
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.321430118602881e-05,
+      "loss": 16.1821,
+      "step": 10477
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.321094253804915e-05,
+      "loss": 16.5301,
+      "step": 10478
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.320758362188227e-05,
+      "loss": 15.9489,
+      "step": 10479
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.320422443755526e-05,
+      "loss": 15.5942,
+      "step": 10480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.320086498509529e-05,
+      "loss": 16.3924,
+      "step": 10481
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.319750526452946e-05,
+      "loss": 17.9394,
+      "step": 10482
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.31941452758849e-05,
+      "loss": 15.38,
+      "step": 10483
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.319078501918873e-05,
+      "loss": 15.3049,
+      "step": 10484
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.318742449446814e-05,
+      "loss": 17.663,
+      "step": 10485
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.318406370175021e-05,
+      "loss": 16.7948,
+      "step": 10486
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.318070264106213e-05,
+      "loss": 17.0079,
+      "step": 10487
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.317734131243098e-05,
+      "loss": 16.4751,
+      "step": 10488
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.317397971588394e-05,
+      "loss": 15.4906,
+      "step": 10489
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.317061785144818e-05,
+      "loss": 16.3574,
+      "step": 10490
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.31672557191508e-05,
+      "loss": 16.7426,
+      "step": 10491
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.316389331901897e-05,
+      "loss": 16.0336,
+      "step": 10492
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.316053065107985e-05,
+      "loss": 15.1131,
+      "step": 10493
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.315716771536058e-05,
+      "loss": 15.5829,
+      "step": 10494
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.315380451188832e-05,
+      "loss": 16.5354,
+      "step": 10495
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.315044104069024e-05,
+      "loss": 15.7381,
+      "step": 10496
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.31470773017935e-05,
+      "loss": 17.8356,
+      "step": 10497
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.314371329522524e-05,
+      "loss": 15.7609,
+      "step": 10498
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.314034902101266e-05,
+      "loss": 15.5796,
+      "step": 10499
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.313698447918289e-05,
+      "loss": 15.2768,
+      "step": 10500
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.313361966976313e-05,
+      "loss": 17.1703,
+      "step": 10501
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.313025459278054e-05,
+      "loss": 16.3466,
+      "step": 10502
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.312688924826228e-05,
+      "loss": 15.3918,
+      "step": 10503
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.312352363623557e-05,
+      "loss": 15.0019,
+      "step": 10504
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.312015775672754e-05,
+      "loss": 16.1719,
+      "step": 10505
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.311679160976539e-05,
+      "loss": 15.9652,
+      "step": 10506
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.31134251953763e-05,
+      "loss": 15.7214,
+      "step": 10507
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.311005851358746e-05,
+      "loss": 19.0063,
+      "step": 10508
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.310669156442606e-05,
+      "loss": 17.4393,
+      "step": 10509
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.310332434791927e-05,
+      "loss": 16.6996,
+      "step": 10510
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.309995686409429e-05,
+      "loss": 16.4971,
+      "step": 10511
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.309658911297834e-05,
+      "loss": 14.8297,
+      "step": 10512
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.309322109459856e-05,
+      "loss": 15.845,
+      "step": 10513
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.308985280898221e-05,
+      "loss": 17.1936,
+      "step": 10514
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.308648425615644e-05,
+      "loss": 15.8077,
+      "step": 10515
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.308311543614848e-05,
+      "loss": 15.4921,
+      "step": 10516
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.307974634898553e-05,
+      "loss": 14.9205,
+      "step": 10517
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.307637699469477e-05,
+      "loss": 15.1772,
+      "step": 10518
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.307300737330346e-05,
+      "loss": 14.6564,
+      "step": 10519
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.306963748483876e-05,
+      "loss": 17.1781,
+      "step": 10520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.306626732932791e-05,
+      "loss": 16.17,
+      "step": 10521
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.306289690679811e-05,
+      "loss": 16.7853,
+      "step": 10522
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.30595262172766e-05,
+      "loss": 14.5171,
+      "step": 10523
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.305615526079058e-05,
+      "loss": 17.2464,
+      "step": 10524
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.305278403736729e-05,
+      "loss": 15.4306,
+      "step": 10525
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.304941254703391e-05,
+      "loss": 17.2436,
+      "step": 10526
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.304604078981772e-05,
+      "loss": 16.3997,
+      "step": 10527
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.304266876574592e-05,
+      "loss": 14.8976,
+      "step": 10528
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.303929647484573e-05,
+      "loss": 15.8696,
+      "step": 10529
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.30359239171444e-05,
+      "loss": 15.6997,
+      "step": 10530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.303255109266918e-05,
+      "loss": 17.4871,
+      "step": 10531
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.302917800144728e-05,
+      "loss": 15.872,
+      "step": 10532
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.302580464350592e-05,
+      "loss": 18.2219,
+      "step": 10533
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.302243101887239e-05,
+      "loss": 15.3321,
+      "step": 10534
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.30190571275739e-05,
+      "loss": 17.1925,
+      "step": 10535
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.301568296963769e-05,
+      "loss": 14.0671,
+      "step": 10536
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.301230854509103e-05,
+      "loss": 15.7377,
+      "step": 10537
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.300893385396116e-05,
+      "loss": 15.1663,
+      "step": 10538
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.300555889627535e-05,
+      "loss": 16.8554,
+      "step": 10539
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.300218367206081e-05,
+      "loss": 16.0725,
+      "step": 10540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.299880818134484e-05,
+      "loss": 14.4452,
+      "step": 10541
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.299543242415467e-05,
+      "loss": 15.3825,
+      "step": 10542
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.299205640051756e-05,
+      "loss": 15.0389,
+      "step": 10543
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.29886801104608e-05,
+      "loss": 16.9495,
+      "step": 10544
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.298530355401164e-05,
+      "loss": 18.9047,
+      "step": 10545
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.298192673119734e-05,
+      "loss": 14.893,
+      "step": 10546
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.297854964204514e-05,
+      "loss": 14.7205,
+      "step": 10547
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.297517228658238e-05,
+      "loss": 15.8544,
+      "step": 10548
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.297179466483628e-05,
+      "loss": 16.7611,
+      "step": 10549
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.296841677683414e-05,
+      "loss": 16.3082,
+      "step": 10550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.29650386226032e-05,
+      "loss": 15.7726,
+      "step": 10551
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.29616602021708e-05,
+      "loss": 16.7896,
+      "step": 10552
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.295828151556417e-05,
+      "loss": 14.2505,
+      "step": 10553
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.295490256281061e-05,
+      "loss": 16.0678,
+      "step": 10554
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.295152334393743e-05,
+      "loss": 15.0525,
+      "step": 10555
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.294814385897189e-05,
+      "loss": 14.1105,
+      "step": 10556
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.294476410794128e-05,
+      "loss": 17.1573,
+      "step": 10557
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.29413840908729e-05,
+      "loss": 17.3248,
+      "step": 10558
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.293800380779404e-05,
+      "loss": 14.364,
+      "step": 10559
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.293462325873201e-05,
+      "loss": 18.4097,
+      "step": 10560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.29312424437141e-05,
+      "loss": 15.4737,
+      "step": 10561
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.29278613627676e-05,
+      "loss": 16.6975,
+      "step": 10562
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.292448001591984e-05,
+      "loss": 13.5415,
+      "step": 10563
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.292109840319811e-05,
+      "loss": 15.4418,
+      "step": 10564
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.29177165246297e-05,
+      "loss": 15.9751,
+      "step": 10565
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.291433438024195e-05,
+      "loss": 16.1368,
+      "step": 10566
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.291095197006214e-05,
+      "loss": 17.7991,
+      "step": 10567
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.290756929411761e-05,
+      "loss": 16.8042,
+      "step": 10568
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.290418635243567e-05,
+      "loss": 15.4579,
+      "step": 10569
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.290080314504363e-05,
+      "loss": 15.5957,
+      "step": 10570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.289741967196882e-05,
+      "loss": 17.1493,
+      "step": 10571
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.289403593323857e-05,
+      "loss": 14.7665,
+      "step": 10572
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.28906519288802e-05,
+      "loss": 16.5735,
+      "step": 10573
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.288726765892101e-05,
+      "loss": 16.297,
+      "step": 10574
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.288388312338835e-05,
+      "loss": 16.2485,
+      "step": 10575
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.288049832230956e-05,
+      "loss": 14.7021,
+      "step": 10576
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.287711325571195e-05,
+      "loss": 16.325,
+      "step": 10577
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.287372792362287e-05,
+      "loss": 15.1369,
+      "step": 10578
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.287034232606966e-05,
+      "loss": 16.5696,
+      "step": 10579
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.286695646307965e-05,
+      "loss": 16.6029,
+      "step": 10580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.28635703346802e-05,
+      "loss": 14.9253,
+      "step": 10581
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.286018394089863e-05,
+      "loss": 17.3808,
+      "step": 10582
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.28567972817623e-05,
+      "loss": 15.6624,
+      "step": 10583
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.285341035729856e-05,
+      "loss": 14.6759,
+      "step": 10584
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.285002316753474e-05,
+      "loss": 16.7,
+      "step": 10585
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.284663571249823e-05,
+      "loss": 14.5242,
+      "step": 10586
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.284324799221634e-05,
+      "loss": 15.5547,
+      "step": 10587
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.283986000671646e-05,
+      "loss": 16.5924,
+      "step": 10588
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.283647175602594e-05,
+      "loss": 15.661,
+      "step": 10589
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.283308324017215e-05,
+      "loss": 16.2707,
+      "step": 10590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.282969445918242e-05,
+      "loss": 16.3902,
+      "step": 10591
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.282630541308416e-05,
+      "loss": 16.27,
+      "step": 10592
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.28229161019047e-05,
+      "loss": 16.2387,
+      "step": 10593
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.281952652567143e-05,
+      "loss": 16.2747,
+      "step": 10594
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.281613668441172e-05,
+      "loss": 14.4405,
+      "step": 10595
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.281274657815293e-05,
+      "loss": 17.7933,
+      "step": 10596
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.280935620692245e-05,
+      "loss": 16.9364,
+      "step": 10597
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.280596557074767e-05,
+      "loss": 17.1045,
+      "step": 10598
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.280257466965593e-05,
+      "loss": 14.8808,
+      "step": 10599
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.279918350367466e-05,
+      "loss": 15.88,
+      "step": 10600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.27957920728312e-05,
+      "loss": 15.9662,
+      "step": 10601
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.279240037715297e-05,
+      "loss": 15.9232,
+      "step": 10602
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.278900841666734e-05,
+      "loss": 14.8425,
+      "step": 10603
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.278561619140172e-05,
+      "loss": 15.4899,
+      "step": 10604
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.278222370138349e-05,
+      "loss": 16.7017,
+      "step": 10605
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.277883094664005e-05,
+      "loss": 15.4355,
+      "step": 10606
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.277543792719878e-05,
+      "loss": 17.2639,
+      "step": 10607
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.27720446430871e-05,
+      "loss": 16.2448,
+      "step": 10608
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.276865109433241e-05,
+      "loss": 14.9439,
+      "step": 10609
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.276525728096212e-05,
+      "loss": 15.2694,
+      "step": 10610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.276186320300361e-05,
+      "loss": 14.3866,
+      "step": 10611
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.27584688604843e-05,
+      "loss": 16.0725,
+      "step": 10612
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.275507425343161e-05,
+      "loss": 15.9346,
+      "step": 10613
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.275167938187294e-05,
+      "loss": 14.3604,
+      "step": 10614
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.274828424583572e-05,
+      "loss": 16.7333,
+      "step": 10615
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.274488884534736e-05,
+      "loss": 15.4303,
+      "step": 10616
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.274149318043526e-05,
+      "loss": 15.7442,
+      "step": 10617
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.273809725112686e-05,
+      "loss": 15.5383,
+      "step": 10618
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.273470105744958e-05,
+      "loss": 16.1513,
+      "step": 10619
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.273130459943086e-05,
+      "loss": 16.4702,
+      "step": 10620
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.27279078770981e-05,
+      "loss": 17.09,
+      "step": 10621
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.272451089047874e-05,
+      "loss": 16.3102,
+      "step": 10622
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.272111363960022e-05,
+      "loss": 14.5894,
+      "step": 10623
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.271771612448995e-05,
+      "loss": 15.6473,
+      "step": 10624
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.27143183451754e-05,
+      "loss": 15.5065,
+      "step": 10625
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.2710920301684e-05,
+      "loss": 14.9095,
+      "step": 10626
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.270752199404316e-05,
+      "loss": 15.1198,
+      "step": 10627
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.270412342228036e-05,
+      "loss": 16.824,
+      "step": 10628
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.270072458642301e-05,
+      "loss": 16.1392,
+      "step": 10629
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.269732548649858e-05,
+      "loss": 15.5923,
+      "step": 10630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.26939261225345e-05,
+      "loss": 16.0042,
+      "step": 10631
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.269052649455826e-05,
+      "loss": 17.4537,
+      "step": 10632
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.268712660259728e-05,
+      "loss": 17.6886,
+      "step": 10633
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.268372644667901e-05,
+      "loss": 14.3594,
+      "step": 10634
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.268032602683093e-05,
+      "loss": 17.9588,
+      "step": 10635
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.267692534308047e-05,
+      "loss": 15.9311,
+      "step": 10636
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.267352439545513e-05,
+      "loss": 13.8781,
+      "step": 10637
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.267012318398234e-05,
+      "loss": 15.4118,
+      "step": 10638
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.266672170868958e-05,
+      "loss": 15.4137,
+      "step": 10639
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.266331996960431e-05,
+      "loss": 15.9288,
+      "step": 10640
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.265991796675402e-05,
+      "loss": 17.8574,
+      "step": 10641
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.265651570016617e-05,
+      "loss": 15.384,
+      "step": 10642
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.265311316986821e-05,
+      "loss": 16.6046,
+      "step": 10643
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.264971037588767e-05,
+      "loss": 14.8823,
+      "step": 10644
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.264630731825198e-05,
+      "loss": 15.3543,
+      "step": 10645
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.264290399698863e-05,
+      "loss": 16.7773,
+      "step": 10646
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.263950041212511e-05,
+      "loss": 17.6996,
+      "step": 10647
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.263609656368893e-05,
+      "loss": 14.5214,
+      "step": 10648
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.263269245170752e-05,
+      "loss": 15.1227,
+      "step": 10649
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.262928807620843e-05,
+      "loss": 14.8961,
+      "step": 10650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.26258834372191e-05,
+      "loss": 12.7364,
+      "step": 10651
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.262247853476708e-05,
+      "loss": 14.3998,
+      "step": 10652
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.26190733688798e-05,
+      "loss": 17.4764,
+      "step": 10653
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.26156679395848e-05,
+      "loss": 15.6667,
+      "step": 10654
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.261226224690956e-05,
+      "loss": 15.9982,
+      "step": 10655
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.26088562908816e-05,
+      "loss": 15.1308,
+      "step": 10656
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.260545007152843e-05,
+      "loss": 15.8049,
+      "step": 10657
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.260204358887754e-05,
+      "loss": 14.7213,
+      "step": 10658
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.25986368429564e-05,
+      "loss": 14.5893,
+      "step": 10659
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.25952298337926e-05,
+      "loss": 16.1086,
+      "step": 10660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.25918225614136e-05,
+      "loss": 16.9034,
+      "step": 10661
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.258841502584692e-05,
+      "loss": 14.8398,
+      "step": 10662
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.258500722712008e-05,
+      "loss": 15.8079,
+      "step": 10663
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.258159916526062e-05,
+      "loss": 16.5104,
+      "step": 10664
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.257819084029603e-05,
+      "loss": 16.6556,
+      "step": 10665
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.257478225225384e-05,
+      "loss": 16.943,
+      "step": 10666
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.25713734011616e-05,
+      "loss": 15.008,
+      "step": 10667
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.25679642870468e-05,
+      "loss": 16.3501,
+      "step": 10668
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.256455490993702e-05,
+      "loss": 16.0766,
+      "step": 10669
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.256114526985972e-05,
+      "loss": 14.793,
+      "step": 10670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.255773536684251e-05,
+      "loss": 16.499,
+      "step": 10671
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.255432520091288e-05,
+      "loss": 13.6687,
+      "step": 10672
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.255091477209836e-05,
+      "loss": 16.508,
+      "step": 10673
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.254750408042652e-05,
+      "loss": 15.5042,
+      "step": 10674
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.25440931259249e-05,
+      "loss": 16.0023,
+      "step": 10675
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.254068190862103e-05,
+      "loss": 14.5561,
+      "step": 10676
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.253727042854245e-05,
+      "loss": 16.823,
+      "step": 10677
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.253385868571672e-05,
+      "loss": 17.5044,
+      "step": 10678
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.25304466801714e-05,
+      "loss": 14.1825,
+      "step": 10679
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.252703441193404e-05,
+      "loss": 16.9767,
+      "step": 10680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.252362188103218e-05,
+      "loss": 16.71,
+      "step": 10681
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.252020908749339e-05,
+      "loss": 15.3854,
+      "step": 10682
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.251679603134522e-05,
+      "loss": 17.0598,
+      "step": 10683
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.251338271261523e-05,
+      "loss": 17.5963,
+      "step": 10684
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.2509969131331e-05,
+      "loss": 16.287,
+      "step": 10685
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.250655528752009e-05,
+      "loss": 13.7009,
+      "step": 10686
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.250314118121005e-05,
+      "loss": 15.5379,
+      "step": 10687
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.249972681242846e-05,
+      "loss": 17.671,
+      "step": 10688
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.24963121812029e-05,
+      "loss": 16.6448,
+      "step": 10689
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.249289728756095e-05,
+      "loss": 17.4817,
+      "step": 10690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.248948213153018e-05,
+      "loss": 16.0496,
+      "step": 10691
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.248606671313814e-05,
+      "loss": 17.517,
+      "step": 10692
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.248265103241244e-05,
+      "loss": 15.9779,
+      "step": 10693
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.247923508938066e-05,
+      "loss": 15.4923,
+      "step": 10694
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.247581888407037e-05,
+      "loss": 15.0542,
+      "step": 10695
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.247240241650918e-05,
+      "loss": 14.5676,
+      "step": 10696
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.246898568672468e-05,
+      "loss": 15.8516,
+      "step": 10697
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.246556869474442e-05,
+      "loss": 14.5251,
+      "step": 10698
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.246215144059604e-05,
+      "loss": 15.2774,
+      "step": 10699
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.24587339243071e-05,
+      "loss": 18.6552,
+      "step": 10700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.245531614590523e-05,
+      "loss": 17.0793,
+      "step": 10701
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.2451898105418e-05,
+      "loss": 15.1504,
+      "step": 10702
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.244847980287302e-05,
+      "loss": 16.0058,
+      "step": 10703
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.244506123829792e-05,
+      "loss": 16.0258,
+      "step": 10704
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.244164241172026e-05,
+      "loss": 14.7708,
+      "step": 10705
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.243822332316768e-05,
+      "loss": 15.8076,
+      "step": 10706
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.243480397266778e-05,
+      "loss": 16.4197,
+      "step": 10707
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.243138436024819e-05,
+      "loss": 15.2298,
+      "step": 10708
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.242796448593649e-05,
+      "loss": 16.1761,
+      "step": 10709
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.242454434976032e-05,
+      "loss": 15.7963,
+      "step": 10710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.242112395174729e-05,
+      "loss": 16.2431,
+      "step": 10711
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.241770329192502e-05,
+      "loss": 17.807,
+      "step": 10712
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.241428237032115e-05,
+      "loss": 18.0849,
+      "step": 10713
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.241086118696329e-05,
+      "loss": 16.294,
+      "step": 10714
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.240743974187906e-05,
+      "loss": 16.2628,
+      "step": 10715
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.24040180350961e-05,
+      "loss": 15.9737,
+      "step": 10716
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.240059606664203e-05,
+      "loss": 14.5918,
+      "step": 10717
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.23971738365445e-05,
+      "loss": 18.0365,
+      "step": 10718
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.239375134483114e-05,
+      "loss": 15.1432,
+      "step": 10719
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.239032859152958e-05,
+      "loss": 14.5602,
+      "step": 10720
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.238690557666748e-05,
+      "loss": 14.7305,
+      "step": 10721
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.238348230027245e-05,
+      "loss": 16.1916,
+      "step": 10722
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.238005876237216e-05,
+      "loss": 15.8588,
+      "step": 10723
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.237663496299424e-05,
+      "loss": 15.2942,
+      "step": 10724
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.237321090216634e-05,
+      "loss": 15.1826,
+      "step": 10725
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.236978657991612e-05,
+      "loss": 15.2894,
+      "step": 10726
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.236636199627122e-05,
+      "loss": 16.7304,
+      "step": 10727
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.236293715125931e-05,
+      "loss": 16.327,
+      "step": 10728
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.235951204490804e-05,
+      "loss": 17.8078,
+      "step": 10729
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.235608667724506e-05,
+      "loss": 15.9716,
+      "step": 10730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.235266104829804e-05,
+      "loss": 15.6874,
+      "step": 10731
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.234923515809462e-05,
+      "loss": 15.2065,
+      "step": 10732
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.234580900666251e-05,
+      "loss": 15.3908,
+      "step": 10733
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.234238259402935e-05,
+      "loss": 15.5424,
+      "step": 10734
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.23389559202228e-05,
+      "loss": 15.168,
+      "step": 10735
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.233552898527055e-05,
+      "loss": 15.2888,
+      "step": 10736
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.233210178920026e-05,
+      "loss": 16.2776,
+      "step": 10737
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.232867433203962e-05,
+      "loss": 15.8372,
+      "step": 10738
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.23252466138163e-05,
+      "loss": 16.3747,
+      "step": 10739
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.232181863455797e-05,
+      "loss": 17.1528,
+      "step": 10740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.231839039429233e-05,
+      "loss": 15.6661,
+      "step": 10741
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.231496189304704e-05,
+      "loss": 14.426,
+      "step": 10742
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.231153313084982e-05,
+      "loss": 16.4541,
+      "step": 10743
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.230810410772834e-05,
+      "loss": 15.938,
+      "step": 10744
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.230467482371027e-05,
+      "loss": 15.8801,
+      "step": 10745
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.230124527882333e-05,
+      "loss": 14.7099,
+      "step": 10746
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.229781547309521e-05,
+      "loss": 16.5516,
+      "step": 10747
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.22943854065536e-05,
+      "loss": 15.1809,
+      "step": 10748
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.22909550792262e-05,
+      "loss": 15.4044,
+      "step": 10749
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.228752449114071e-05,
+      "loss": 16.6329,
+      "step": 10750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.228409364232484e-05,
+      "loss": 16.2665,
+      "step": 10751
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.228066253280629e-05,
+      "loss": 17.5946,
+      "step": 10752
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.227723116261277e-05,
+      "loss": 15.7034,
+      "step": 10753
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.227379953177197e-05,
+      "loss": 18.1809,
+      "step": 10754
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.227036764031162e-05,
+      "loss": 18.8798,
+      "step": 10755
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.226693548825945e-05,
+      "loss": 14.9269,
+      "step": 10756
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.226350307564315e-05,
+      "loss": 14.795,
+      "step": 10757
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.226007040249044e-05,
+      "loss": 13.8251,
+      "step": 10758
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.225663746882903e-05,
+      "loss": 13.4161,
+      "step": 10759
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.225320427468668e-05,
+      "loss": 17.0327,
+      "step": 10760
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.224977082009108e-05,
+      "loss": 17.2193,
+      "step": 10761
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.224633710506997e-05,
+      "loss": 17.0031,
+      "step": 10762
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.224290312965108e-05,
+      "loss": 16.5781,
+      "step": 10763
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.223946889386212e-05,
+      "loss": 16.1887,
+      "step": 10764
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.223603439773084e-05,
+      "loss": 14.5177,
+      "step": 10765
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.223259964128499e-05,
+      "loss": 15.4818,
+      "step": 10766
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.222916462455226e-05,
+      "loss": 18.6334,
+      "step": 10767
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.222572934756043e-05,
+      "loss": 16.2706,
+      "step": 10768
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.222229381033722e-05,
+      "loss": 13.875,
+      "step": 10769
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.22188580129104e-05,
+      "loss": 15.6907,
+      "step": 10770
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.221542195530767e-05,
+      "loss": 15.678,
+      "step": 10771
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.221198563755682e-05,
+      "loss": 16.7754,
+      "step": 10772
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.220854905968559e-05,
+      "loss": 15.1275,
+      "step": 10773
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.220511222172169e-05,
+      "loss": 16.2985,
+      "step": 10774
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.220167512369293e-05,
+      "loss": 15.2311,
+      "step": 10775
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.219823776562704e-05,
+      "loss": 14.9314,
+      "step": 10776
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.219480014755178e-05,
+      "loss": 15.6782,
+      "step": 10777
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.219136226949489e-05,
+      "loss": 17.3151,
+      "step": 10778
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.218792413148418e-05,
+      "loss": 16.5553,
+      "step": 10779
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.218448573354736e-05,
+      "loss": 15.4275,
+      "step": 10780
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.218104707571223e-05,
+      "loss": 16.904,
+      "step": 10781
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.217760815800656e-05,
+      "loss": 17.9377,
+      "step": 10782
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.217416898045809e-05,
+      "loss": 15.6348,
+      "step": 10783
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.217072954309462e-05,
+      "loss": 16.105,
+      "step": 10784
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.21672898459439e-05,
+      "loss": 15.4091,
+      "step": 10785
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.216384988903376e-05,
+      "loss": 15.7673,
+      "step": 10786
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.216040967239192e-05,
+      "loss": 15.3689,
+      "step": 10787
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.215696919604618e-05,
+      "loss": 16.2044,
+      "step": 10788
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.215352846002431e-05,
+      "loss": 16.0088,
+      "step": 10789
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.215008746435414e-05,
+      "loss": 14.1681,
+      "step": 10790
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.21466462090634e-05,
+      "loss": 17.2648,
+      "step": 10791
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.21432046941799e-05,
+      "loss": 15.3344,
+      "step": 10792
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.213976291973145e-05,
+      "loss": 14.4961,
+      "step": 10793
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.213632088574583e-05,
+      "loss": 16.3507,
+      "step": 10794
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.213287859225084e-05,
+      "loss": 15.4418,
+      "step": 10795
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.212943603927425e-05,
+      "loss": 15.4738,
+      "step": 10796
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.212599322684391e-05,
+      "loss": 15.1908,
+      "step": 10797
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.212255015498758e-05,
+      "loss": 15.3188,
+      "step": 10798
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.211910682373308e-05,
+      "loss": 19.2695,
+      "step": 10799
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.21156632331082e-05,
+      "loss": 17.0838,
+      "step": 10800
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.211221938314077e-05,
+      "loss": 14.6072,
+      "step": 10801
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.210877527385859e-05,
+      "loss": 16.3478,
+      "step": 10802
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.210533090528947e-05,
+      "loss": 17.8899,
+      "step": 10803
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.210188627746123e-05,
+      "loss": 15.3014,
+      "step": 10804
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.209844139040169e-05,
+      "loss": 14.987,
+      "step": 10805
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.209499624413865e-05,
+      "loss": 15.6885,
+      "step": 10806
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.209155083869994e-05,
+      "loss": 15.5643,
+      "step": 10807
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.208810517411339e-05,
+      "loss": 15.7813,
+      "step": 10808
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.208465925040682e-05,
+      "loss": 17.4019,
+      "step": 10809
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.208121306760805e-05,
+      "loss": 16.4702,
+      "step": 10810
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.207776662574493e-05,
+      "loss": 16.6084,
+      "step": 10811
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.207431992484527e-05,
+      "loss": 16.8088,
+      "step": 10812
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.20708729649369e-05,
+      "loss": 15.0153,
+      "step": 10813
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.206742574604767e-05,
+      "loss": 15.1882,
+      "step": 10814
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.206397826820542e-05,
+      "loss": 14.3109,
+      "step": 10815
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.206053053143796e-05,
+      "loss": 17.205,
+      "step": 10816
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.205708253577318e-05,
+      "loss": 16.9501,
+      "step": 10817
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.205363428123887e-05,
+      "loss": 15.6262,
+      "step": 10818
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.205018576786291e-05,
+      "loss": 15.6795,
+      "step": 10819
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.204673699567313e-05,
+      "loss": 16.0055,
+      "step": 10820
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.20432879646974e-05,
+      "loss": 16.8377,
+      "step": 10821
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.203983867496356e-05,
+      "loss": 14.4032,
+      "step": 10822
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.203638912649947e-05,
+      "loss": 15.7987,
+      "step": 10823
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.203293931933298e-05,
+      "loss": 14.8368,
+      "step": 10824
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.202948925349194e-05,
+      "loss": 16.83,
+      "step": 10825
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.202603892900422e-05,
+      "loss": 17.7697,
+      "step": 10826
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.202258834589767e-05,
+      "loss": 14.8987,
+      "step": 10827
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.201913750420018e-05,
+      "loss": 16.9712,
+      "step": 10828
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.20156864039396e-05,
+      "loss": 15.1223,
+      "step": 10829
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.201223504514379e-05,
+      "loss": 15.3489,
+      "step": 10830
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.200878342784064e-05,
+      "loss": 17.224,
+      "step": 10831
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.200533155205802e-05,
+      "loss": 16.189,
+      "step": 10832
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.200187941782379e-05,
+      "loss": 13.9528,
+      "step": 10833
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.199842702516583e-05,
+      "loss": 15.9776,
+      "step": 10834
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.199497437411204e-05,
+      "loss": 14.2966,
+      "step": 10835
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.199152146469027e-05,
+      "loss": 14.4672,
+      "step": 10836
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.198806829692842e-05,
+      "loss": 14.1592,
+      "step": 10837
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.198461487085438e-05,
+      "loss": 15.7252,
+      "step": 10838
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.198116118649604e-05,
+      "loss": 15.4197,
+      "step": 10839
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.197770724388126e-05,
+      "loss": 16.5545,
+      "step": 10840
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.197425304303796e-05,
+      "loss": 17.4925,
+      "step": 10841
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.197079858399403e-05,
+      "loss": 14.6896,
+      "step": 10842
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.196734386677736e-05,
+      "loss": 14.3888,
+      "step": 10843
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.196388889141584e-05,
+      "loss": 16.6052,
+      "step": 10844
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.19604336579374e-05,
+      "loss": 17.5663,
+      "step": 10845
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.19569781663699e-05,
+      "loss": 15.1863,
+      "step": 10846
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.195352241674127e-05,
+      "loss": 16.5795,
+      "step": 10847
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.195006640907942e-05,
+      "loss": 16.8217,
+      "step": 10848
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.194661014341224e-05,
+      "loss": 16.7863,
+      "step": 10849
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.194315361976765e-05,
+      "loss": 16.7577,
+      "step": 10850
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.193969683817358e-05,
+      "loss": 15.4597,
+      "step": 10851
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.19362397986579e-05,
+      "loss": 14.1178,
+      "step": 10852
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.193278250124856e-05,
+      "loss": 18.0102,
+      "step": 10853
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.192932494597347e-05,
+      "loss": 16.4667,
+      "step": 10854
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.192586713286056e-05,
+      "loss": 17.6239,
+      "step": 10855
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.192240906193776e-05,
+      "loss": 15.6968,
+      "step": 10856
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.191895073323294e-05,
+      "loss": 14.1201,
+      "step": 10857
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.19154921467741e-05,
+      "loss": 16.6611,
+      "step": 10858
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.191203330258912e-05,
+      "loss": 15.1061,
+      "step": 10859
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.190857420070596e-05,
+      "loss": 17.4557,
+      "step": 10860
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.190511484115253e-05,
+      "loss": 16.8449,
+      "step": 10861
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.190165522395676e-05,
+      "loss": 16.2323,
+      "step": 10862
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.189819534914662e-05,
+      "loss": 14.8333,
+      "step": 10863
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.189473521675004e-05,
+      "loss": 15.5613,
+      "step": 10864
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.189127482679495e-05,
+      "loss": 14.9176,
+      "step": 10865
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.188781417930928e-05,
+      "loss": 13.5435,
+      "step": 10866
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.1884353274321e-05,
+      "loss": 15.8097,
+      "step": 10867
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.188089211185807e-05,
+      "loss": 14.7825,
+      "step": 10868
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.18774306919484e-05,
+      "loss": 16.5159,
+      "step": 10869
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.187396901461996e-05,
+      "loss": 16.5363,
+      "step": 10870
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.187050707990072e-05,
+      "loss": 15.7405,
+      "step": 10871
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.186704488781862e-05,
+      "loss": 16.3485,
+      "step": 10872
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.186358243840161e-05,
+      "loss": 15.5175,
+      "step": 10873
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.186011973167766e-05,
+      "loss": 15.4612,
+      "step": 10874
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.185665676767475e-05,
+      "loss": 15.3051,
+      "step": 10875
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.18531935464208e-05,
+      "loss": 17.2091,
+      "step": 10876
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.184973006794383e-05,
+      "loss": 18.1612,
+      "step": 10877
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.184626633227178e-05,
+      "loss": 14.652,
+      "step": 10878
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.18428023394326e-05,
+      "loss": 16.1705,
+      "step": 10879
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.18393380894543e-05,
+      "loss": 16.7425,
+      "step": 10880
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.183587358236484e-05,
+      "loss": 14.8747,
+      "step": 10881
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.183240881819219e-05,
+      "loss": 15.9349,
+      "step": 10882
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.182894379696434e-05,
+      "loss": 13.5969,
+      "step": 10883
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.182547851870926e-05,
+      "loss": 16.9667,
+      "step": 10884
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.182201298345496e-05,
+      "loss": 18.6571,
+      "step": 10885
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.181854719122939e-05,
+      "loss": 15.2679,
+      "step": 10886
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.181508114206056e-05,
+      "loss": 16.0499,
+      "step": 10887
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.181161483597644e-05,
+      "loss": 16.591,
+      "step": 10888
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.180814827300503e-05,
+      "loss": 15.5974,
+      "step": 10889
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.180468145317433e-05,
+      "loss": 15.9553,
+      "step": 10890
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.180121437651234e-05,
+      "loss": 16.5289,
+      "step": 10891
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.179774704304704e-05,
+      "loss": 15.5492,
+      "step": 10892
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.179427945280645e-05,
+      "loss": 17.8986,
+      "step": 10893
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.179081160581855e-05,
+      "loss": 12.6607,
+      "step": 10894
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.178734350211135e-05,
+      "loss": 14.0163,
+      "step": 10895
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.178387514171288e-05,
+      "loss": 17.6289,
+      "step": 10896
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.17804065246511e-05,
+      "loss": 13.4605,
+      "step": 10897
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.177693765095408e-05,
+      "loss": 19.7652,
+      "step": 10898
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.177346852064977e-05,
+      "loss": 15.2881,
+      "step": 10899
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.176999913376624e-05,
+      "loss": 15.1823,
+      "step": 10900
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.176652949033148e-05,
+      "loss": 16.8448,
+      "step": 10901
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.17630595903735e-05,
+      "loss": 16.1854,
+      "step": 10902
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.175958943392032e-05,
+      "loss": 16.5602,
+      "step": 10903
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.175611902099997e-05,
+      "loss": 15.4419,
+      "step": 10904
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.17526483516405e-05,
+      "loss": 16.514,
+      "step": 10905
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.174917742586989e-05,
+      "loss": 15.5918,
+      "step": 10906
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.17457062437162e-05,
+      "loss": 14.0914,
+      "step": 10907
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.174223480520747e-05,
+      "loss": 16.8888,
+      "step": 10908
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.173876311037168e-05,
+      "loss": 15.0672,
+      "step": 10909
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.173529115923693e-05,
+      "loss": 15.8027,
+      "step": 10910
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.17318189518312e-05,
+      "loss": 17.5889,
+      "step": 10911
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.172834648818257e-05,
+      "loss": 14.4327,
+      "step": 10912
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.172487376831907e-05,
+      "loss": 16.6807,
+      "step": 10913
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.172140079226874e-05,
+      "loss": 15.119,
+      "step": 10914
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.171792756005962e-05,
+      "loss": 17.3907,
+      "step": 10915
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.171445407171977e-05,
+      "loss": 15.4123,
+      "step": 10916
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.171098032727723e-05,
+      "loss": 16.0152,
+      "step": 10917
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.170750632676006e-05,
+      "loss": 15.6493,
+      "step": 10918
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.170403207019627e-05,
+      "loss": 15.8885,
+      "step": 10919
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.170055755761399e-05,
+      "loss": 15.8139,
+      "step": 10920
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.169708278904122e-05,
+      "loss": 15.6066,
+      "step": 10921
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.169360776450606e-05,
+      "loss": 15.1849,
+      "step": 10922
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.169013248403654e-05,
+      "loss": 16.5911,
+      "step": 10923
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.168665694766073e-05,
+      "loss": 16.2036,
+      "step": 10924
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.168318115540671e-05,
+      "loss": 17.3668,
+      "step": 10925
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.167970510730253e-05,
+      "loss": 17.7821,
+      "step": 10926
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.167622880337628e-05,
+      "loss": 15.2331,
+      "step": 10927
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.1672752243656e-05,
+      "loss": 17.0813,
+      "step": 10928
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.16692754281698e-05,
+      "loss": 15.627,
+      "step": 10929
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.166579835694575e-05,
+      "loss": 16.7075,
+      "step": 10930
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.16623210300119e-05,
+      "loss": 14.8914,
+      "step": 10931
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.165884344739637e-05,
+      "loss": 16.073,
+      "step": 10932
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.16553656091272e-05,
+      "loss": 17.3795,
+      "step": 10933
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.165188751523251e-05,
+      "loss": 16.4423,
+      "step": 10934
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.16484091657404e-05,
+      "loss": 15.2963,
+      "step": 10935
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.16449305606789e-05,
+      "loss": 15.3516,
+      "step": 10936
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.164145170007613e-05,
+      "loss": 16.8954,
+      "step": 10937
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.163797258396019e-05,
+      "loss": 15.66,
+      "step": 10938
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.163449321235918e-05,
+      "loss": 15.9102,
+      "step": 10939
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.163101358530117e-05,
+      "loss": 16.0749,
+      "step": 10940
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.162753370281428e-05,
+      "loss": 17.2144,
+      "step": 10941
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.162405356492661e-05,
+      "loss": 15.2052,
+      "step": 10942
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.162057317166628e-05,
+      "loss": 16.393,
+      "step": 10943
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.161709252306136e-05,
+      "loss": 17.1267,
+      "step": 10944
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.161361161913996e-05,
+      "loss": 16.5551,
+      "step": 10945
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.161013045993023e-05,
+      "loss": 16.4761,
+      "step": 10946
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.160664904546023e-05,
+      "loss": 16.4477,
+      "step": 10947
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.160316737575811e-05,
+      "loss": 15.2493,
+      "step": 10948
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.159968545085195e-05,
+      "loss": 15.7038,
+      "step": 10949
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.159620327076991e-05,
+      "loss": 16.403,
+      "step": 10950
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.159272083554008e-05,
+      "loss": 15.582,
+      "step": 10951
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.15892381451906e-05,
+      "loss": 14.9841,
+      "step": 10952
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.158575519974956e-05,
+      "loss": 15.3145,
+      "step": 10953
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.158227199924513e-05,
+      "loss": 15.9488,
+      "step": 10954
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.157878854370541e-05,
+      "loss": 19.4987,
+      "step": 10955
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.157530483315854e-05,
+      "loss": 15.0933,
+      "step": 10956
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.157182086763265e-05,
+      "loss": 15.7904,
+      "step": 10957
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.156833664715589e-05,
+      "loss": 15.4731,
+      "step": 10958
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.156485217175636e-05,
+      "loss": 15.0399,
+      "step": 10959
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.156136744146221e-05,
+      "loss": 15.0652,
+      "step": 10960
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.155788245630159e-05,
+      "loss": 15.2888,
+      "step": 10961
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.155439721630264e-05,
+      "loss": 14.6808,
+      "step": 10962
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.155091172149351e-05,
+      "loss": 15.8414,
+      "step": 10963
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.154742597190235e-05,
+      "loss": 19.0733,
+      "step": 10964
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.154393996755728e-05,
+      "loss": 14.4988,
+      "step": 10965
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.154045370848648e-05,
+      "loss": 17.1806,
+      "step": 10966
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.153696719471809e-05,
+      "loss": 15.4707,
+      "step": 10967
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.153348042628025e-05,
+      "loss": 15.2727,
+      "step": 10968
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.152999340320116e-05,
+      "loss": 14.6363,
+      "step": 10969
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.152650612550891e-05,
+      "loss": 15.4621,
+      "step": 10970
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.152301859323174e-05,
+      "loss": 16.3266,
+      "step": 10971
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.151953080639775e-05,
+      "loss": 16.0367,
+      "step": 10972
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.151604276503515e-05,
+      "loss": 15.4711,
+      "step": 10973
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.151255446917207e-05,
+      "loss": 15.4059,
+      "step": 10974
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.150906591883671e-05,
+      "loss": 16.5779,
+      "step": 10975
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.150557711405721e-05,
+      "loss": 17.8149,
+      "step": 10976
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.150208805486178e-05,
+      "loss": 16.5668,
+      "step": 10977
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.149859874127856e-05,
+      "loss": 15.7598,
+      "step": 10978
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.149510917333574e-05,
+      "loss": 15.074,
+      "step": 10979
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.149161935106151e-05,
+      "loss": 17.4592,
+      "step": 10980
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.148812927448402e-05,
+      "loss": 16.237,
+      "step": 10981
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.148463894363149e-05,
+      "loss": 17.8244,
+      "step": 10982
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.14811483585321e-05,
+      "loss": 16.8077,
+      "step": 10983
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.147765751921401e-05,
+      "loss": 16.5036,
+      "step": 10984
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.147416642570544e-05,
+      "loss": 16.6076,
+      "step": 10985
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.147067507803456e-05,
+      "loss": 16.945,
+      "step": 10986
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.146718347622957e-05,
+      "loss": 16.7731,
+      "step": 10987
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.146369162031868e-05,
+      "loss": 14.7504,
+      "step": 10988
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.146019951033007e-05,
+      "loss": 15.7477,
+      "step": 10989
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.145670714629195e-05,
+      "loss": 16.2962,
+      "step": 10990
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.145321452823251e-05,
+      "loss": 14.1697,
+      "step": 10991
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.144972165617996e-05,
+      "loss": 15.1429,
+      "step": 10992
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.14462285301625e-05,
+      "loss": 15.2785,
+      "step": 10993
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.144273515020837e-05,
+      "loss": 15.2563,
+      "step": 10994
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.143924151634573e-05,
+      "loss": 16.1763,
+      "step": 10995
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.143574762860282e-05,
+      "loss": 17.9616,
+      "step": 10996
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.143225348700788e-05,
+      "loss": 16.0894,
+      "step": 10997
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.142875909158906e-05,
+      "loss": 16.3874,
+      "step": 10998
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.142526444237463e-05,
+      "loss": 15.8755,
+      "step": 10999
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.142176953939279e-05,
+      "loss": 16.7383,
+      "step": 11000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.141827438267177e-05,
+      "loss": 17.118,
+      "step": 11001
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.14147789722398e-05,
+      "loss": 17.3314,
+      "step": 11002
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.141128330812508e-05,
+      "loss": 17.0806,
+      "step": 11003
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.140778739035588e-05,
+      "loss": 15.7244,
+      "step": 11004
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.14042912189604e-05,
+      "loss": 14.9277,
+      "step": 11005
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.140079479396687e-05,
+      "loss": 16.9654,
+      "step": 11006
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.139729811540353e-05,
+      "loss": 16.2312,
+      "step": 11007
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.139380118329864e-05,
+      "loss": 15.069,
+      "step": 11008
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.13903039976804e-05,
+      "loss": 16.5118,
+      "step": 11009
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.138680655857708e-05,
+      "loss": 15.9737,
+      "step": 11010
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.138330886601693e-05,
+      "loss": 15.2792,
+      "step": 11011
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.137981092002816e-05,
+      "loss": 15.9109,
+      "step": 11012
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.137631272063903e-05,
+      "loss": 16.1544,
+      "step": 11013
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.13728142678778e-05,
+      "loss": 15.576,
+      "step": 11014
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.13693155617727e-05,
+      "loss": 15.4028,
+      "step": 11015
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.136581660235201e-05,
+      "loss": 16.0343,
+      "step": 11016
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.136231738964398e-05,
+      "loss": 16.134,
+      "step": 11017
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.135881792367686e-05,
+      "loss": 15.22,
+      "step": 11018
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.13553182044789e-05,
+      "loss": 16.2867,
+      "step": 11019
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.135181823207837e-05,
+      "loss": 15.7611,
+      "step": 11020
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.134831800650353e-05,
+      "loss": 14.9488,
+      "step": 11021
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.134481752778265e-05,
+      "loss": 16.5919,
+      "step": 11022
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.134131679594402e-05,
+      "loss": 14.9139,
+      "step": 11023
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.133781581101585e-05,
+      "loss": 14.2082,
+      "step": 11024
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.133431457302646e-05,
+      "loss": 15.9405,
+      "step": 11025
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.13308130820041e-05,
+      "loss": 17.0817,
+      "step": 11026
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.132731133797708e-05,
+      "loss": 15.0662,
+      "step": 11027
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.132380934097363e-05,
+      "loss": 15.6785,
+      "step": 11028
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.132030709102205e-05,
+      "loss": 17.0692,
+      "step": 11029
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.131680458815064e-05,
+      "loss": 15.9512,
+      "step": 11030
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.131330183238766e-05,
+      "loss": 15.9608,
+      "step": 11031
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.13097988237614e-05,
+      "loss": 14.8367,
+      "step": 11032
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.130629556230016e-05,
+      "loss": 16.3096,
+      "step": 11033
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.130279204803221e-05,
+      "loss": 16.5711,
+      "step": 11034
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.129928828098587e-05,
+      "loss": 16.0595,
+      "step": 11035
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.12957842611894e-05,
+      "loss": 16.2018,
+      "step": 11036
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.129227998867113e-05,
+      "loss": 15.0825,
+      "step": 11037
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.128877546345933e-05,
+      "loss": 15.5822,
+      "step": 11038
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.12852706855823e-05,
+      "loss": 15.3089,
+      "step": 11039
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.128176565506837e-05,
+      "loss": 15.3114,
+      "step": 11040
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.127826037194581e-05,
+      "loss": 16.8558,
+      "step": 11041
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.127475483624294e-05,
+      "loss": 14.6412,
+      "step": 11042
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.127124904798809e-05,
+      "loss": 16.526,
+      "step": 11043
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.126774300720956e-05,
+      "loss": 16.7218,
+      "step": 11044
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.126423671393563e-05,
+      "loss": 17.4011,
+      "step": 11045
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.126073016819463e-05,
+      "loss": 17.4822,
+      "step": 11046
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.12572233700149e-05,
+      "loss": 17.0796,
+      "step": 11047
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.125371631942473e-05,
+      "loss": 14.3339,
+      "step": 11048
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.125020901645246e-05,
+      "loss": 15.8442,
+      "step": 11049
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.124670146112641e-05,
+      "loss": 16.7407,
+      "step": 11050
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.12431936534749e-05,
+      "loss": 17.549,
+      "step": 11051
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.123968559352623e-05,
+      "loss": 17.4446,
+      "step": 11052
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.12361772813088e-05,
+      "loss": 15.0832,
+      "step": 11053
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.123266871685085e-05,
+      "loss": 16.5991,
+      "step": 11054
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.122915990018077e-05,
+      "loss": 16.4817,
+      "step": 11055
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.12256508313269e-05,
+      "loss": 15.0466,
+      "step": 11056
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.122214151031753e-05,
+      "loss": 17.289,
+      "step": 11057
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.121863193718104e-05,
+      "loss": 14.9287,
+      "step": 11058
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.121512211194577e-05,
+      "loss": 15.2927,
+      "step": 11059
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.121161203464004e-05,
+      "loss": 16.198,
+      "step": 11060
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.12081017052922e-05,
+      "loss": 16.0063,
+      "step": 11061
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.120459112393062e-05,
+      "loss": 15.8722,
+      "step": 11062
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.120108029058361e-05,
+      "loss": 16.0875,
+      "step": 11063
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.119756920527955e-05,
+      "loss": 16.0881,
+      "step": 11064
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.119405786804679e-05,
+      "loss": 17.9076,
+      "step": 11065
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.119054627891368e-05,
+      "loss": 14.9235,
+      "step": 11066
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.118703443790859e-05,
+      "loss": 16.6335,
+      "step": 11067
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.118352234505984e-05,
+      "loss": 16.5072,
+      "step": 11068
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.118001000039583e-05,
+      "loss": 16.0674,
+      "step": 11069
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.11764974039449e-05,
+      "loss": 15.2314,
+      "step": 11070
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.117298455573543e-05,
+      "loss": 14.8435,
+      "step": 11071
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.11694714557958e-05,
+      "loss": 17.9127,
+      "step": 11072
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.116595810415435e-05,
+      "loss": 15.1755,
+      "step": 11073
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.116244450083945e-05,
+      "loss": 16.6783,
+      "step": 11074
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.11589306458795e-05,
+      "loss": 15.9382,
+      "step": 11075
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.115541653930286e-05,
+      "loss": 15.7396,
+      "step": 11076
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.115190218113791e-05,
+      "loss": 15.9404,
+      "step": 11077
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.114838757141304e-05,
+      "loss": 16.5982,
+      "step": 11078
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.11448727101566e-05,
+      "loss": 17.8102,
+      "step": 11079
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.114135759739701e-05,
+      "loss": 15.0253,
+      "step": 11080
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.113784223316263e-05,
+      "loss": 17.0389,
+      "step": 11081
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.113432661748185e-05,
+      "loss": 13.9513,
+      "step": 11082
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.113081075038307e-05,
+      "loss": 16.9994,
+      "step": 11083
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.112729463189468e-05,
+      "loss": 14.3934,
+      "step": 11084
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.112377826204508e-05,
+      "loss": 16.6069,
+      "step": 11085
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.112026164086266e-05,
+      "loss": 15.9476,
+      "step": 11086
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.111674476837581e-05,
+      "loss": 16.4839,
+      "step": 11087
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.111322764461291e-05,
+      "loss": 14.9648,
+      "step": 11088
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.110971026960241e-05,
+      "loss": 14.7548,
+      "step": 11089
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.11061926433727e-05,
+      "loss": 15.8954,
+      "step": 11090
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.110267476595215e-05,
+      "loss": 16.0262,
+      "step": 11091
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.109915663736919e-05,
+      "loss": 17.5656,
+      "step": 11092
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.109563825765225e-05,
+      "loss": 16.5404,
+      "step": 11093
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.109211962682971e-05,
+      "loss": 16.8987,
+      "step": 11094
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.108860074493e-05,
+      "loss": 17.2945,
+      "step": 11095
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.108508161198155e-05,
+      "loss": 15.2823,
+      "step": 11096
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.108156222801275e-05,
+      "loss": 13.8726,
+      "step": 11097
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.107804259305205e-05,
+      "loss": 17.4082,
+      "step": 11098
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.107452270712783e-05,
+      "loss": 15.6346,
+      "step": 11099
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.107100257026857e-05,
+      "loss": 15.8263,
+      "step": 11100
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.106748218250263e-05,
+      "loss": 18.0124,
+      "step": 11101
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.106396154385848e-05,
+      "loss": 16.9957,
+      "step": 11102
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.106044065436456e-05,
+      "loss": 15.2142,
+      "step": 11103
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.105691951404928e-05,
+      "loss": 15.624,
+      "step": 11104
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.105339812294107e-05,
+      "loss": 17.2415,
+      "step": 11105
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.104987648106838e-05,
+      "loss": 14.7222,
+      "step": 11106
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.104635458845964e-05,
+      "loss": 15.1235,
+      "step": 11107
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.10428324451433e-05,
+      "loss": 16.7893,
+      "step": 11108
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.103931005114778e-05,
+      "loss": 16.0462,
+      "step": 11109
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.103578740650156e-05,
+      "loss": 15.761,
+      "step": 11110
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.103226451123306e-05,
+      "loss": 14.03,
+      "step": 11111
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.102874136537073e-05,
+      "loss": 15.3618,
+      "step": 11112
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.102521796894303e-05,
+      "loss": 15.4856,
+      "step": 11113
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.102169432197842e-05,
+      "loss": 13.8742,
+      "step": 11114
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.101817042450531e-05,
+      "loss": 15.7086,
+      "step": 11115
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.10146462765522e-05,
+      "loss": 16.3066,
+      "step": 11116
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.101112187814755e-05,
+      "loss": 17.3054,
+      "step": 11117
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.100759722931981e-05,
+      "loss": 15.5124,
+      "step": 11118
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.100407233009742e-05,
+      "loss": 16.2052,
+      "step": 11119
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.100054718050887e-05,
+      "loss": 16.801,
+      "step": 11120
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.099702178058262e-05,
+      "loss": 15.0157,
+      "step": 11121
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.099349613034715e-05,
+      "loss": 16.958,
+      "step": 11122
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.098997022983092e-05,
+      "loss": 15.7536,
+      "step": 11123
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.09864440790624e-05,
+      "loss": 15.3445,
+      "step": 11124
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.098291767807004e-05,
+      "loss": 14.7254,
+      "step": 11125
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.097939102688238e-05,
+      "loss": 17.0095,
+      "step": 11126
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.097586412552785e-05,
+      "loss": 15.8866,
+      "step": 11127
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.097233697403494e-05,
+      "loss": 15.0143,
+      "step": 11128
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.096880957243214e-05,
+      "loss": 16.9535,
+      "step": 11129
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.096528192074793e-05,
+      "loss": 15.4491,
+      "step": 11130
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.096175401901079e-05,
+      "loss": 17.547,
+      "step": 11131
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.095822586724921e-05,
+      "loss": 17.2042,
+      "step": 11132
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.095469746549172e-05,
+      "loss": 16.8529,
+      "step": 11133
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.095116881376675e-05,
+      "loss": 16.7535,
+      "step": 11134
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.094763991210282e-05,
+      "loss": 16.5571,
+      "step": 11135
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.094411076052845e-05,
+      "loss": 15.7121,
+      "step": 11136
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.09405813590721e-05,
+      "loss": 14.3947,
+      "step": 11137
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.09370517077623e-05,
+      "loss": 16.3868,
+      "step": 11138
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.093352180662755e-05,
+      "loss": 15.6258,
+      "step": 11139
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.092999165569634e-05,
+      "loss": 17.8295,
+      "step": 11140
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.092646125499719e-05,
+      "loss": 15.1863,
+      "step": 11141
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.09229306045586e-05,
+      "loss": 18.111,
+      "step": 11142
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.09193997044091e-05,
+      "loss": 16.854,
+      "step": 11143
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.091586855457717e-05,
+      "loss": 16.9232,
+      "step": 11144
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.091233715509135e-05,
+      "loss": 15.3893,
+      "step": 11145
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.090880550598015e-05,
+      "loss": 16.6071,
+      "step": 11146
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.090527360727209e-05,
+      "loss": 16.5532,
+      "step": 11147
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.09017414589957e-05,
+      "loss": 13.536,
+      "step": 11148
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.089820906117947e-05,
+      "loss": 16.2748,
+      "step": 11149
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.089467641385196e-05,
+      "loss": 15.1081,
+      "step": 11150
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.089114351704168e-05,
+      "loss": 15.3116,
+      "step": 11151
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.088761037077718e-05,
+      "loss": 15.2692,
+      "step": 11152
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.088407697508697e-05,
+      "loss": 16.9849,
+      "step": 11153
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.088054332999957e-05,
+      "loss": 18.2551,
+      "step": 11154
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.087700943554357e-05,
+      "loss": 15.4653,
+      "step": 11155
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.087347529174743e-05,
+      "loss": 15.7997,
+      "step": 11156
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.086994089863975e-05,
+      "loss": 17.3318,
+      "step": 11157
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.086640625624905e-05,
+      "loss": 15.9279,
+      "step": 11158
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.086287136460388e-05,
+      "loss": 15.2276,
+      "step": 11159
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.085933622373278e-05,
+      "loss": 16.8065,
+      "step": 11160
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.085580083366429e-05,
+      "loss": 15.4566,
+      "step": 11161
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.085226519442697e-05,
+      "loss": 16.3709,
+      "step": 11162
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.084872930604936e-05,
+      "loss": 16.2031,
+      "step": 11163
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.084519316856003e-05,
+      "loss": 16.2761,
+      "step": 11164
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.084165678198751e-05,
+      "loss": 16.7442,
+      "step": 11165
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.083812014636039e-05,
+      "loss": 15.2945,
+      "step": 11166
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.083458326170721e-05,
+      "loss": 16.8941,
+      "step": 11167
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.083104612805653e-05,
+      "loss": 15.2704,
+      "step": 11168
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.082750874543692e-05,
+      "loss": 13.4327,
+      "step": 11169
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.082397111387694e-05,
+      "loss": 15.5773,
+      "step": 11170
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.082043323340515e-05,
+      "loss": 14.6887,
+      "step": 11171
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.081689510405015e-05,
+      "loss": 16.2538,
+      "step": 11172
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.081335672584048e-05,
+      "loss": 16.3164,
+      "step": 11173
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.080981809880473e-05,
+      "loss": 15.591,
+      "step": 11174
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.080627922297146e-05,
+      "loss": 16.9535,
+      "step": 11175
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.080274009836926e-05,
+      "loss": 13.2783,
+      "step": 11176
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.07992007250267e-05,
+      "loss": 15.5745,
+      "step": 11177
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.079566110297236e-05,
+      "loss": 16.6577,
+      "step": 11178
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.079212123223486e-05,
+      "loss": 16.2499,
+      "step": 11179
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.078858111284273e-05,
+      "loss": 13.6657,
+      "step": 11180
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.078504074482458e-05,
+      "loss": 14.7991,
+      "step": 11181
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.078150012820901e-05,
+      "loss": 15.3024,
+      "step": 11182
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.077795926302461e-05,
+      "loss": 15.2799,
+      "step": 11183
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.077441814929996e-05,
+      "loss": 16.1125,
+      "step": 11184
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.077087678706368e-05,
+      "loss": 16.2104,
+      "step": 11185
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.076733517634433e-05,
+      "loss": 15.6521,
+      "step": 11186
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.076379331717053e-05,
+      "loss": 14.8206,
+      "step": 11187
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.076025120957089e-05,
+      "loss": 17.6036,
+      "step": 11188
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.075670885357401e-05,
+      "loss": 15.0523,
+      "step": 11189
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.075316624920848e-05,
+      "loss": 15.0292,
+      "step": 11190
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.074962339650291e-05,
+      "loss": 15.8545,
+      "step": 11191
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.074608029548593e-05,
+      "loss": 15.71,
+      "step": 11192
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.074253694618614e-05,
+      "loss": 15.6154,
+      "step": 11193
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.073899334863215e-05,
+      "loss": 17.0297,
+      "step": 11194
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.073544950285259e-05,
+      "loss": 14.9812,
+      "step": 11195
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.073190540887606e-05,
+      "loss": 14.8122,
+      "step": 11196
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.072836106673116e-05,
+      "loss": 17.0022,
+      "step": 11197
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.072481647644656e-05,
+      "loss": 16.9332,
+      "step": 11198
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.072127163805085e-05,
+      "loss": 16.8421,
+      "step": 11199
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.071772655157267e-05,
+      "loss": 15.0175,
+      "step": 11200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.071418121704064e-05,
+      "loss": 17.117,
+      "step": 11201
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.07106356344834e-05,
+      "loss": 17.0458,
+      "step": 11202
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.070708980392956e-05,
+      "loss": 15.6631,
+      "step": 11203
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.070354372540779e-05,
+      "loss": 15.7484,
+      "step": 11204
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.069999739894667e-05,
+      "loss": 16.4,
+      "step": 11205
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.069645082457488e-05,
+      "loss": 15.019,
+      "step": 11206
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.069290400232109e-05,
+      "loss": 14.8821,
+      "step": 11207
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.068935693221385e-05,
+      "loss": 16.3563,
+      "step": 11208
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.068580961428187e-05,
+      "loss": 15.4604,
+      "step": 11209
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.068226204855379e-05,
+      "loss": 17.0455,
+      "step": 11210
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.067871423505824e-05,
+      "loss": 16.0965,
+      "step": 11211
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.067516617382388e-05,
+      "loss": 16.266,
+      "step": 11212
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.067161786487935e-05,
+      "loss": 15.6158,
+      "step": 11213
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.066806930825333e-05,
+      "loss": 17.2438,
+      "step": 11214
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.066452050397446e-05,
+      "loss": 14.8061,
+      "step": 11215
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.066097145207139e-05,
+      "loss": 15.9966,
+      "step": 11216
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.065742215257278e-05,
+      "loss": 15.0051,
+      "step": 11217
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.06538726055073e-05,
+      "loss": 15.7279,
+      "step": 11218
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.065032281090361e-05,
+      "loss": 15.8427,
+      "step": 11219
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.064677276879038e-05,
+      "loss": 14.6581,
+      "step": 11220
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.064322247919627e-05,
+      "loss": 17.4634,
+      "step": 11221
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.063967194214994e-05,
+      "loss": 16.7461,
+      "step": 11222
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.063612115768009e-05,
+      "loss": 16.5822,
+      "step": 11223
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.063257012581538e-05,
+      "loss": 16.7931,
+      "step": 11224
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.062901884658448e-05,
+      "loss": 13.951,
+      "step": 11225
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.062546732001607e-05,
+      "loss": 13.842,
+      "step": 11226
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.062191554613882e-05,
+      "loss": 15.984,
+      "step": 11227
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.061836352498145e-05,
+      "loss": 16.0865,
+      "step": 11228
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.061481125657259e-05,
+      "loss": 15.3476,
+      "step": 11229
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.061125874094095e-05,
+      "loss": 17.3675,
+      "step": 11230
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.060770597811525e-05,
+      "loss": 17.1482,
+      "step": 11231
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.060415296812412e-05,
+      "loss": 16.2644,
+      "step": 11232
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 8.060059971099627e-05,
+      "loss": 16.3975,
+      "step": 11233
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.059704620676042e-05,
+      "loss": 14.6935,
+      "step": 11234
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.059349245544526e-05,
+      "loss": 16.2411,
+      "step": 11235
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.058993845707947e-05,
+      "loss": 16.3524,
+      "step": 11236
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.058638421169175e-05,
+      "loss": 16.6959,
+      "step": 11237
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.05828297193108e-05,
+      "loss": 17.016,
+      "step": 11238
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.057927497996534e-05,
+      "loss": 15.6719,
+      "step": 11239
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.057571999368408e-05,
+      "loss": 16.5605,
+      "step": 11240
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.057216476049569e-05,
+      "loss": 16.8588,
+      "step": 11241
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.05686092804289e-05,
+      "loss": 17.5289,
+      "step": 11242
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.056505355351245e-05,
+      "loss": 15.8664,
+      "step": 11243
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.056149757977501e-05,
+      "loss": 16.6495,
+      "step": 11244
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.055794135924531e-05,
+      "loss": 16.5578,
+      "step": 11245
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.055438489195208e-05,
+      "loss": 15.1111,
+      "step": 11246
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.055082817792403e-05,
+      "loss": 17.9794,
+      "step": 11247
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.054727121718987e-05,
+      "loss": 14.4281,
+      "step": 11248
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.054371400977836e-05,
+      "loss": 15.5973,
+      "step": 11249
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.054015655571819e-05,
+      "loss": 16.2587,
+      "step": 11250
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.05365988550381e-05,
+      "loss": 14.8223,
+      "step": 11251
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.053304090776681e-05,
+      "loss": 16.1579,
+      "step": 11252
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.052948271393306e-05,
+      "loss": 15.7305,
+      "step": 11253
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.052592427356558e-05,
+      "loss": 16.7561,
+      "step": 11254
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.05223655866931e-05,
+      "loss": 14.2412,
+      "step": 11255
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.051880665334439e-05,
+      "loss": 15.6268,
+      "step": 11256
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.051524747354813e-05,
+      "loss": 16.3394,
+      "step": 11257
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.051168804733313e-05,
+      "loss": 15.8721,
+      "step": 11258
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.050812837472808e-05,
+      "loss": 14.7874,
+      "step": 11259
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.050456845576175e-05,
+      "loss": 15.6073,
+      "step": 11260
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.050100829046287e-05,
+      "loss": 16.0686,
+      "step": 11261
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.049744787886022e-05,
+      "loss": 14.5504,
+      "step": 11262
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.049388722098252e-05,
+      "loss": 16.2268,
+      "step": 11263
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.049032631685853e-05,
+      "loss": 14.826,
+      "step": 11264
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.048676516651702e-05,
+      "loss": 16.1019,
+      "step": 11265
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.048320376998673e-05,
+      "loss": 16.5409,
+      "step": 11266
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.047964212729644e-05,
+      "loss": 15.3247,
+      "step": 11267
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.047608023847489e-05,
+      "loss": 17.9314,
+      "step": 11268
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.047251810355084e-05,
+      "loss": 18.2495,
+      "step": 11269
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.046895572255309e-05,
+      "loss": 14.8683,
+      "step": 11270
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.046539309551035e-05,
+      "loss": 15.8965,
+      "step": 11271
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.046183022245146e-05,
+      "loss": 16.0332,
+      "step": 11272
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.045826710340514e-05,
+      "loss": 16.8993,
+      "step": 11273
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.045470373840017e-05,
+      "loss": 15.8679,
+      "step": 11274
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.045114012746533e-05,
+      "loss": 14.7245,
+      "step": 11275
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.04475762706294e-05,
+      "loss": 16.2326,
+      "step": 11276
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.044401216792116e-05,
+      "loss": 14.1406,
+      "step": 11277
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.044044781936942e-05,
+      "loss": 15.3476,
+      "step": 11278
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.043688322500289e-05,
+      "loss": 15.5161,
+      "step": 11279
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.043331838485042e-05,
+      "loss": 14.7579,
+      "step": 11280
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.042975329894076e-05,
+      "loss": 16.1204,
+      "step": 11281
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.042618796730273e-05,
+      "loss": 16.303,
+      "step": 11282
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.04226223899651e-05,
+      "loss": 20.2013,
+      "step": 11283
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.041905656695666e-05,
+      "loss": 16.9611,
+      "step": 11284
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.04154904983062e-05,
+      "loss": 16.3584,
+      "step": 11285
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.041192418404257e-05,
+      "loss": 17.85,
+      "step": 11286
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.040835762419449e-05,
+      "loss": 17.8721,
+      "step": 11287
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.040479081879082e-05,
+      "loss": 17.1218,
+      "step": 11288
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.040122376786034e-05,
+      "loss": 15.9382,
+      "step": 11289
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.039765647143186e-05,
+      "loss": 15.5616,
+      "step": 11290
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.039408892953418e-05,
+      "loss": 16.8818,
+      "step": 11291
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.039052114219612e-05,
+      "loss": 16.147,
+      "step": 11292
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.038695310944647e-05,
+      "loss": 15.0189,
+      "step": 11293
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.038338483131407e-05,
+      "loss": 15.8456,
+      "step": 11294
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.037981630782771e-05,
+      "loss": 16.3968,
+      "step": 11295
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.037624753901623e-05,
+      "loss": 17.0759,
+      "step": 11296
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.037267852490841e-05,
+      "loss": 16.6549,
+      "step": 11297
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.036910926553312e-05,
+      "loss": 16.9944,
+      "step": 11298
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.036553976091917e-05,
+      "loss": 15.5233,
+      "step": 11299
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.036197001109536e-05,
+      "loss": 14.7756,
+      "step": 11300
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.035840001609054e-05,
+      "loss": 16.5814,
+      "step": 11301
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.035482977593352e-05,
+      "loss": 15.7917,
+      "step": 11302
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.035125929065315e-05,
+      "loss": 15.9656,
+      "step": 11303
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.034768856027826e-05,
+      "loss": 17.4822,
+      "step": 11304
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.034411758483766e-05,
+      "loss": 16.3999,
+      "step": 11305
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.034054636436021e-05,
+      "loss": 14.5451,
+      "step": 11306
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.033697489887475e-05,
+      "loss": 16.5619,
+      "step": 11307
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.033340318841011e-05,
+      "loss": 15.7777,
+      "step": 11308
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.032983123299515e-05,
+      "loss": 15.9722,
+      "step": 11309
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.032625903265869e-05,
+      "loss": 15.219,
+      "step": 11310
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.032268658742959e-05,
+      "loss": 14.836,
+      "step": 11311
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.031911389733669e-05,
+      "loss": 14.8059,
+      "step": 11312
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.031554096240884e-05,
+      "loss": 13.848,
+      "step": 11313
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.031196778267492e-05,
+      "loss": 15.4697,
+      "step": 11314
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.030839435816374e-05,
+      "loss": 15.8269,
+      "step": 11315
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.030482068890419e-05,
+      "loss": 14.5312,
+      "step": 11316
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.030124677492512e-05,
+      "loss": 16.538,
+      "step": 11317
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.029767261625538e-05,
+      "loss": 16.7143,
+      "step": 11318
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.029409821292386e-05,
+      "loss": 14.5346,
+      "step": 11319
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.029052356495939e-05,
+      "loss": 15.9282,
+      "step": 11320
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.028694867239084e-05,
+      "loss": 15.9737,
+      "step": 11321
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.02833735352471e-05,
+      "loss": 15.4055,
+      "step": 11322
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.027979815355704e-05,
+      "loss": 14.3237,
+      "step": 11323
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.027622252734952e-05,
+      "loss": 13.8294,
+      "step": 11324
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.02726466566534e-05,
+      "loss": 15.499,
+      "step": 11325
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.02690705414976e-05,
+      "loss": 14.4217,
+      "step": 11326
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.026549418191093e-05,
+      "loss": 17.1805,
+      "step": 11327
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.026191757792234e-05,
+      "loss": 14.7002,
+      "step": 11328
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.025834072956067e-05,
+      "loss": 14.6391,
+      "step": 11329
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.025476363685482e-05,
+      "loss": 18.1493,
+      "step": 11330
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.025118629983367e-05,
+      "loss": 16.8248,
+      "step": 11331
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.024760871852611e-05,
+      "loss": 16.747,
+      "step": 11332
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.024403089296104e-05,
+      "loss": 16.3445,
+      "step": 11333
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.024045282316733e-05,
+      "loss": 17.4753,
+      "step": 11334
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.023687450917389e-05,
+      "loss": 15.2524,
+      "step": 11335
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.023329595100962e-05,
+      "loss": 16.3144,
+      "step": 11336
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.022971714870337e-05,
+      "loss": 15.1668,
+      "step": 11337
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.022613810228413e-05,
+      "loss": 17.4729,
+      "step": 11338
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.022255881178071e-05,
+      "loss": 15.9637,
+      "step": 11339
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.021897927722208e-05,
+      "loss": 16.3881,
+      "step": 11340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.021539949863712e-05,
+      "loss": 17.0787,
+      "step": 11341
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.021181947605473e-05,
+      "loss": 16.3458,
+      "step": 11342
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.020823920950384e-05,
+      "loss": 15.5721,
+      "step": 11343
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.020465869901332e-05,
+      "loss": 15.6478,
+      "step": 11344
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.020107794461214e-05,
+      "loss": 16.2492,
+      "step": 11345
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.019749694632918e-05,
+      "loss": 16.5274,
+      "step": 11346
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.019391570419336e-05,
+      "loss": 17.635,
+      "step": 11347
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.01903342182336e-05,
+      "loss": 15.0279,
+      "step": 11348
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.018675248847886e-05,
+      "loss": 15.1029,
+      "step": 11349
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.0183170514958e-05,
+      "loss": 16.1784,
+      "step": 11350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.017958829769998e-05,
+      "loss": 16.8744,
+      "step": 11351
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.017600583673372e-05,
+      "loss": 18.2412,
+      "step": 11352
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.017242313208817e-05,
+      "loss": 15.6064,
+      "step": 11353
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.016884018379223e-05,
+      "loss": 16.9876,
+      "step": 11354
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.016525699187485e-05,
+      "loss": 14.9513,
+      "step": 11355
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.016167355636496e-05,
+      "loss": 16.5942,
+      "step": 11356
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.01580898772915e-05,
+      "loss": 15.9017,
+      "step": 11357
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.015450595468341e-05,
+      "loss": 16.6407,
+      "step": 11358
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.015092178856962e-05,
+      "loss": 16.1628,
+      "step": 11359
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.014733737897909e-05,
+      "loss": 14.7873,
+      "step": 11360
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.014375272594076e-05,
+      "loss": 15.3651,
+      "step": 11361
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.014016782948358e-05,
+      "loss": 18.4594,
+      "step": 11362
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.013658268963648e-05,
+      "loss": 14.5331,
+      "step": 11363
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.013299730642842e-05,
+      "loss": 15.908,
+      "step": 11364
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.012941167988839e-05,
+      "loss": 16.3467,
+      "step": 11365
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.012582581004529e-05,
+      "loss": 16.4374,
+      "step": 11366
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.01222396969281e-05,
+      "loss": 15.1583,
+      "step": 11367
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.011865334056577e-05,
+      "loss": 15.4071,
+      "step": 11368
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.011506674098728e-05,
+      "loss": 14.9626,
+      "step": 11369
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.011147989822158e-05,
+      "loss": 15.2021,
+      "step": 11370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.010789281229763e-05,
+      "loss": 14.6795,
+      "step": 11371
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.010430548324442e-05,
+      "loss": 15.77,
+      "step": 11372
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.010071791109088e-05,
+      "loss": 15.5758,
+      "step": 11373
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.009713009586601e-05,
+      "loss": 15.3711,
+      "step": 11374
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.009354203759877e-05,
+      "loss": 16.1392,
+      "step": 11375
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.008995373631816e-05,
+      "loss": 15.8158,
+      "step": 11376
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.008636519205313e-05,
+      "loss": 14.5096,
+      "step": 11377
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.008277640483266e-05,
+      "loss": 15.9412,
+      "step": 11378
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.007918737468573e-05,
+      "loss": 16.2292,
+      "step": 11379
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.007559810164133e-05,
+      "loss": 15.3219,
+      "step": 11380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.007200858572843e-05,
+      "loss": 16.9868,
+      "step": 11381
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.006841882697604e-05,
+      "loss": 16.4405,
+      "step": 11382
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.006482882541314e-05,
+      "loss": 14.6217,
+      "step": 11383
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.006123858106871e-05,
+      "loss": 13.7113,
+      "step": 11384
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.005764809397175e-05,
+      "loss": 16.3632,
+      "step": 11385
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.005405736415126e-05,
+      "loss": 17.4507,
+      "step": 11386
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.005046639163623e-05,
+      "loss": 16.853,
+      "step": 11387
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.004687517645564e-05,
+      "loss": 17.1821,
+      "step": 11388
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.004328371863851e-05,
+      "loss": 14.7202,
+      "step": 11389
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.003969201821385e-05,
+      "loss": 17.0869,
+      "step": 11390
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.003610007521066e-05,
+      "loss": 15.5355,
+      "step": 11391
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.003250788965791e-05,
+      "loss": 16.809,
+      "step": 11392
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.002891546158466e-05,
+      "loss": 15.4684,
+      "step": 11393
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.002532279101989e-05,
+      "loss": 15.9034,
+      "step": 11394
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.002172987799261e-05,
+      "loss": 15.8271,
+      "step": 11395
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.001813672253185e-05,
+      "loss": 17.0908,
+      "step": 11396
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.001454332466661e-05,
+      "loss": 16.8163,
+      "step": 11397
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.001094968442592e-05,
+      "loss": 15.676,
+      "step": 11398
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.00073558018388e-05,
+      "loss": 17.4498,
+      "step": 11399
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.000376167693427e-05,
+      "loss": 18.0062,
+      "step": 11400
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.000016730974133e-05,
+      "loss": 15.59,
+      "step": 11401
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.999657270028905e-05,
+      "loss": 17.8593,
+      "step": 11402
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.999297784860642e-05,
+      "loss": 15.5262,
+      "step": 11403
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.998938275472247e-05,
+      "loss": 17.3192,
+      "step": 11404
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.998578741866627e-05,
+      "loss": 16.1267,
+      "step": 11405
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.99821918404668e-05,
+      "loss": 15.6193,
+      "step": 11406
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.997859602015314e-05,
+      "loss": 16.0501,
+      "step": 11407
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.99749999577543e-05,
+      "loss": 15.9401,
+      "step": 11408
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.997140365329934e-05,
+      "loss": 17.4108,
+      "step": 11409
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.996780710681728e-05,
+      "loss": 14.7745,
+      "step": 11410
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.996421031833716e-05,
+      "loss": 17.7088,
+      "step": 11411
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.996061328788806e-05,
+      "loss": 15.9603,
+      "step": 11412
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.995701601549899e-05,
+      "loss": 14.9327,
+      "step": 11413
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.995341850119903e-05,
+      "loss": 16.4228,
+      "step": 11414
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.99498207450172e-05,
+      "loss": 17.172,
+      "step": 11415
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.994622274698258e-05,
+      "loss": 16.9238,
+      "step": 11416
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.99426245071242e-05,
+      "loss": 15.4137,
+      "step": 11417
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.993902602547113e-05,
+      "loss": 15.4601,
+      "step": 11418
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.993542730205244e-05,
+      "loss": 16.2699,
+      "step": 11419
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.993182833689717e-05,
+      "loss": 16.139,
+      "step": 11420
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.992822913003437e-05,
+      "loss": 17.3727,
+      "step": 11421
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.992462968149315e-05,
+      "loss": 15.3009,
+      "step": 11422
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.992102999130256e-05,
+      "loss": 15.911,
+      "step": 11423
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.991743005949164e-05,
+      "loss": 16.3288,
+      "step": 11424
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.991382988608949e-05,
+      "loss": 15.6329,
+      "step": 11425
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.991022947112517e-05,
+      "loss": 15.7697,
+      "step": 11426
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.990662881462776e-05,
+      "loss": 15.5278,
+      "step": 11427
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.990302791662632e-05,
+      "loss": 14.8801,
+      "step": 11428
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.989942677714997e-05,
+      "loss": 17.1283,
+      "step": 11429
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.989582539622775e-05,
+      "loss": 15.2624,
+      "step": 11430
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.989222377388874e-05,
+      "loss": 16.1791,
+      "step": 11431
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.988862191016205e-05,
+      "loss": 15.9656,
+      "step": 11432
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.988501980507674e-05,
+      "loss": 15.5577,
+      "step": 11433
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.988141745866193e-05,
+      "loss": 16.1588,
+      "step": 11434
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.987781487094669e-05,
+      "loss": 15.1353,
+      "step": 11435
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.987421204196011e-05,
+      "loss": 17.1231,
+      "step": 11436
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.987060897173128e-05,
+      "loss": 16.026,
+      "step": 11437
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.986700566028933e-05,
+      "loss": 16.4482,
+      "step": 11438
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.986340210766329e-05,
+      "loss": 16.4325,
+      "step": 11439
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.985979831388233e-05,
+      "loss": 15.5807,
+      "step": 11440
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.985619427897553e-05,
+      "loss": 16.028,
+      "step": 11441
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.985259000297196e-05,
+      "loss": 17.3041,
+      "step": 11442
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.984898548590077e-05,
+      "loss": 15.1938,
+      "step": 11443
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.984538072779104e-05,
+      "loss": 15.191,
+      "step": 11444
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.98417757286719e-05,
+      "loss": 16.0267,
+      "step": 11445
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.983817048857243e-05,
+      "loss": 15.894,
+      "step": 11446
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.983456500752178e-05,
+      "loss": 15.7551,
+      "step": 11447
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.983095928554904e-05,
+      "loss": 15.2881,
+      "step": 11448
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.982735332268335e-05,
+      "loss": 15.0052,
+      "step": 11449
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.982374711895379e-05,
+      "loss": 17.8008,
+      "step": 11450
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.982014067438953e-05,
+      "loss": 12.3315,
+      "step": 11451
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.981653398901965e-05,
+      "loss": 17.6329,
+      "step": 11452
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.981292706287329e-05,
+      "loss": 16.9354,
+      "step": 11453
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.98093198959796e-05,
+      "loss": 17.2682,
+      "step": 11454
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.980571248836767e-05,
+      "loss": 16.0775,
+      "step": 11455
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.980210484006666e-05,
+      "loss": 17.203,
+      "step": 11456
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.979849695110568e-05,
+      "loss": 14.708,
+      "step": 11457
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.97948888215139e-05,
+      "loss": 16.9053,
+      "step": 11458
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.979128045132042e-05,
+      "loss": 17.0568,
+      "step": 11459
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.97876718405544e-05,
+      "loss": 14.611,
+      "step": 11460
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.978406298924497e-05,
+      "loss": 16.1861,
+      "step": 11461
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.978045389742126e-05,
+      "loss": 16.0573,
+      "step": 11462
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.977684456511246e-05,
+      "loss": 17.7751,
+      "step": 11463
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.977323499234765e-05,
+      "loss": 14.3401,
+      "step": 11464
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.976962517915604e-05,
+      "loss": 16.8168,
+      "step": 11465
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.976601512556676e-05,
+      "loss": 15.6244,
+      "step": 11466
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.976240483160892e-05,
+      "loss": 17.4637,
+      "step": 11467
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.975879429731174e-05,
+      "loss": 15.8089,
+      "step": 11468
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.975518352270434e-05,
+      "loss": 17.3368,
+      "step": 11469
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.975157250781591e-05,
+      "loss": 14.9702,
+      "step": 11470
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.974796125267557e-05,
+      "loss": 14.759,
+      "step": 11471
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.974434975731248e-05,
+      "loss": 17.0775,
+      "step": 11472
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.974073802175583e-05,
+      "loss": 17.3419,
+      "step": 11473
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.973712604603477e-05,
+      "loss": 16.2309,
+      "step": 11474
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.973351383017848e-05,
+      "loss": 16.3714,
+      "step": 11475
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.972990137421612e-05,
+      "loss": 16.804,
+      "step": 11476
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.972628867817688e-05,
+      "loss": 15.0046,
+      "step": 11477
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.972267574208991e-05,
+      "loss": 14.5386,
+      "step": 11478
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.971906256598438e-05,
+      "loss": 15.2423,
+      "step": 11479
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.971544914988949e-05,
+      "loss": 15.4062,
+      "step": 11480
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.97118354938344e-05,
+      "loss": 15.8101,
+      "step": 11481
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.970822159784832e-05,
+      "loss": 18.4483,
+      "step": 11482
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.970460746196042e-05,
+      "loss": 16.5627,
+      "step": 11483
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.970099308619988e-05,
+      "loss": 15.9925,
+      "step": 11484
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.969737847059587e-05,
+      "loss": 15.0261,
+      "step": 11485
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.96937636151776e-05,
+      "loss": 16.5858,
+      "step": 11486
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.969014851997425e-05,
+      "loss": 15.3974,
+      "step": 11487
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.968653318501503e-05,
+      "loss": 15.1354,
+      "step": 11488
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.968291761032912e-05,
+      "loss": 15.5462,
+      "step": 11489
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.967930179594573e-05,
+      "loss": 15.4028,
+      "step": 11490
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.967568574189404e-05,
+      "loss": 17.1364,
+      "step": 11491
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.967206944820327e-05,
+      "loss": 18.8067,
+      "step": 11492
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.966845291490261e-05,
+      "loss": 15.3559,
+      "step": 11493
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.966483614202128e-05,
+      "loss": 17.2646,
+      "step": 11494
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.966121912958844e-05,
+      "loss": 16.5255,
+      "step": 11495
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.965760187763336e-05,
+      "loss": 16.4779,
+      "step": 11496
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.965398438618521e-05,
+      "loss": 15.5958,
+      "step": 11497
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.965036665527323e-05,
+      "loss": 15.4508,
+      "step": 11498
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.964674868492661e-05,
+      "loss": 15.3265,
+      "step": 11499
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.964313047517458e-05,
+      "loss": 14.9713,
+      "step": 11500
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.963951202604633e-05,
+      "loss": 16.0218,
+      "step": 11501
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.963589333757113e-05,
+      "loss": 14.3418,
+      "step": 11502
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.963227440977816e-05,
+      "loss": 16.4436,
+      "step": 11503
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.962865524269667e-05,
+      "loss": 16.0385,
+      "step": 11504
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.962503583635587e-05,
+      "loss": 14.6685,
+      "step": 11505
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.962141619078498e-05,
+      "loss": 15.2453,
+      "step": 11506
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.961779630601326e-05,
+      "loss": 14.0746,
+      "step": 11507
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.961417618206992e-05,
+      "loss": 14.6453,
+      "step": 11508
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.961055581898419e-05,
+      "loss": 15.4594,
+      "step": 11509
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.960693521678532e-05,
+      "loss": 16.0479,
+      "step": 11510
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.960331437550252e-05,
+      "loss": 17.104,
+      "step": 11511
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.959969329516507e-05,
+      "loss": 15.8132,
+      "step": 11512
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.959607197580219e-05,
+      "loss": 14.8937,
+      "step": 11513
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.959245041744312e-05,
+      "loss": 15.4928,
+      "step": 11514
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.958882862011711e-05,
+      "loss": 14.2163,
+      "step": 11515
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.95852065838534e-05,
+      "loss": 14.7571,
+      "step": 11516
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.958158430868124e-05,
+      "loss": 14.8592,
+      "step": 11517
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.957796179462989e-05,
+      "loss": 16.318,
+      "step": 11518
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.957433904172859e-05,
+      "loss": 17.1329,
+      "step": 11519
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.957071605000661e-05,
+      "loss": 15.5234,
+      "step": 11520
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.95670928194932e-05,
+      "loss": 17.4753,
+      "step": 11521
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.956346935021761e-05,
+      "loss": 15.9667,
+      "step": 11522
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.955984564220912e-05,
+      "loss": 15.5457,
+      "step": 11523
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.955622169549697e-05,
+      "loss": 15.3185,
+      "step": 11524
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.955259751011044e-05,
+      "loss": 14.7027,
+      "step": 11525
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.954897308607879e-05,
+      "loss": 16.2544,
+      "step": 11526
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.954534842343129e-05,
+      "loss": 14.3739,
+      "step": 11527
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.954172352219721e-05,
+      "loss": 14.2623,
+      "step": 11528
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.953809838240583e-05,
+      "loss": 15.2682,
+      "step": 11529
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.95344730040864e-05,
+      "loss": 16.95,
+      "step": 11530
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.953084738726822e-05,
+      "loss": 15.8994,
+      "step": 11531
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.952722153198054e-05,
+      "loss": 15.5918,
+      "step": 11532
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.952359543825268e-05,
+      "loss": 16.361,
+      "step": 11533
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.95199691061139e-05,
+      "loss": 15.6674,
+      "step": 11534
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.951634253559348e-05,
+      "loss": 15.225,
+      "step": 11535
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.95127157267207e-05,
+      "loss": 15.9474,
+      "step": 11536
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.950908867952486e-05,
+      "loss": 15.4626,
+      "step": 11537
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.950546139403525e-05,
+      "loss": 15.9543,
+      "step": 11538
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.950183387028116e-05,
+      "loss": 16.1992,
+      "step": 11539
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.949820610829188e-05,
+      "loss": 16.6948,
+      "step": 11540
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.949457810809672e-05,
+      "loss": 15.8534,
+      "step": 11541
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.949094986972494e-05,
+      "loss": 16.02,
+      "step": 11542
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.948732139320586e-05,
+      "loss": 17.5043,
+      "step": 11543
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.948369267856881e-05,
+      "loss": 17.5932,
+      "step": 11544
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.948006372584306e-05,
+      "loss": 17.6888,
+      "step": 11545
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.94764345350579e-05,
+      "loss": 15.1028,
+      "step": 11546
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.947280510624267e-05,
+      "loss": 15.6141,
+      "step": 11547
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.946917543942667e-05,
+      "loss": 14.9882,
+      "step": 11548
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.946554553463919e-05,
+      "loss": 14.7496,
+      "step": 11549
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.946191539190957e-05,
+      "loss": 17.2772,
+      "step": 11550
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.945828501126713e-05,
+      "loss": 14.2352,
+      "step": 11551
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.945465439274114e-05,
+      "loss": 16.4362,
+      "step": 11552
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.945102353636097e-05,
+      "loss": 15.652,
+      "step": 11553
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.944739244215591e-05,
+      "loss": 17.5493,
+      "step": 11554
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.944376111015528e-05,
+      "loss": 17.5915,
+      "step": 11555
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.944012954038842e-05,
+      "loss": 16.232,
+      "step": 11556
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.943649773288467e-05,
+      "loss": 17.2988,
+      "step": 11557
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.943286568767331e-05,
+      "loss": 15.4858,
+      "step": 11558
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.942923340478372e-05,
+      "loss": 16.2105,
+      "step": 11559
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.94256008842452e-05,
+      "loss": 14.82,
+      "step": 11560
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.94219681260871e-05,
+      "loss": 17.6839,
+      "step": 11561
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.941833513033873e-05,
+      "loss": 16.106,
+      "step": 11562
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.941470189702945e-05,
+      "loss": 13.9037,
+      "step": 11563
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.94110684261886e-05,
+      "loss": 14.7256,
+      "step": 11564
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.940743471784551e-05,
+      "loss": 15.1179,
+      "step": 11565
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.940380077202954e-05,
+      "loss": 16.991,
+      "step": 11566
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.940016658877002e-05,
+      "loss": 15.7709,
+      "step": 11567
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.93965321680963e-05,
+      "loss": 14.5478,
+      "step": 11568
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.939289751003774e-05,
+      "loss": 15.3612,
+      "step": 11569
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.938926261462366e-05,
+      "loss": 15.357,
+      "step": 11570
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.938562748188346e-05,
+      "loss": 16.7034,
+      "step": 11571
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.938199211184644e-05,
+      "loss": 15.4582,
+      "step": 11572
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.937835650454199e-05,
+      "loss": 15.3178,
+      "step": 11573
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.937472065999949e-05,
+      "loss": 15.4587,
+      "step": 11574
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.937108457824825e-05,
+      "loss": 14.6829,
+      "step": 11575
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.936744825931767e-05,
+      "loss": 16.9542,
+      "step": 11576
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.936381170323709e-05,
+      "loss": 14.5359,
+      "step": 11577
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.936017491003589e-05,
+      "loss": 16.1579,
+      "step": 11578
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.935653787974345e-05,
+      "loss": 17.2264,
+      "step": 11579
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.93529006123891e-05,
+      "loss": 15.1761,
+      "step": 11580
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.934926310800225e-05,
+      "loss": 16.7972,
+      "step": 11581
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.934562536661225e-05,
+      "loss": 17.039,
+      "step": 11582
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.93419873882485e-05,
+      "loss": 15.2279,
+      "step": 11583
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.933834917294036e-05,
+      "loss": 17.0021,
+      "step": 11584
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.93347107207172e-05,
+      "loss": 17.6063,
+      "step": 11585
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.933107203160845e-05,
+      "loss": 17.2882,
+      "step": 11586
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.932743310564344e-05,
+      "loss": 15.5746,
+      "step": 11587
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.932379394285157e-05,
+      "loss": 16.0717,
+      "step": 11588
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.932015454326223e-05,
+      "loss": 18.0944,
+      "step": 11589
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.931651490690479e-05,
+      "loss": 14.1415,
+      "step": 11590
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.931287503380869e-05,
+      "loss": 15.4562,
+      "step": 11591
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.930923492400329e-05,
+      "loss": 14.8442,
+      "step": 11592
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.930559457751799e-05,
+      "loss": 14.2558,
+      "step": 11593
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.930195399438217e-05,
+      "loss": 17.8156,
+      "step": 11594
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.929831317462526e-05,
+      "loss": 16.2817,
+      "step": 11595
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.929467211827663e-05,
+      "loss": 14.8361,
+      "step": 11596
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.929103082536571e-05,
+      "loss": 16.1849,
+      "step": 11597
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.92873892959219e-05,
+      "loss": 14.5806,
+      "step": 11598
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.928374752997459e-05,
+      "loss": 17.4371,
+      "step": 11599
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.928010552755318e-05,
+      "loss": 14.3729,
+      "step": 11600
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.927646328868712e-05,
+      "loss": 16.2641,
+      "step": 11601
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.927282081340579e-05,
+      "loss": 14.8682,
+      "step": 11602
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.926917810173861e-05,
+      "loss": 17.3113,
+      "step": 11603
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.9265535153715e-05,
+      "loss": 16.3368,
+      "step": 11604
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.926189196936438e-05,
+      "loss": 14.4428,
+      "step": 11605
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.925824854871616e-05,
+      "loss": 14.9653,
+      "step": 11606
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.925460489179977e-05,
+      "loss": 19.0503,
+      "step": 11607
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.925096099864464e-05,
+      "loss": 15.6281,
+      "step": 11608
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.924731686928017e-05,
+      "loss": 15.0472,
+      "step": 11609
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.92436725037358e-05,
+      "loss": 17.0926,
+      "step": 11610
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.924002790204098e-05,
+      "loss": 16.495,
+      "step": 11611
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.923638306422511e-05,
+      "loss": 15.1771,
+      "step": 11612
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.923273799031765e-05,
+      "loss": 17.2348,
+      "step": 11613
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.922909268034801e-05,
+      "loss": 15.3101,
+      "step": 11614
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.922544713434565e-05,
+      "loss": 16.1893,
+      "step": 11615
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.922180135234e-05,
+      "loss": 14.7045,
+      "step": 11616
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.92181553343605e-05,
+      "loss": 17.2453,
+      "step": 11617
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.921450908043657e-05,
+      "loss": 16.5421,
+      "step": 11618
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.921086259059769e-05,
+      "loss": 14.7683,
+      "step": 11619
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.92072158648733e-05,
+      "loss": 15.9805,
+      "step": 11620
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.920356890329282e-05,
+      "loss": 15.9064,
+      "step": 11621
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.919992170588573e-05,
+      "loss": 16.1184,
+      "step": 11622
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.919627427268148e-05,
+      "loss": 18.2745,
+      "step": 11623
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.919262660370951e-05,
+      "loss": 15.245,
+      "step": 11624
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.918897869899929e-05,
+      "loss": 15.5311,
+      "step": 11625
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.918533055858025e-05,
+      "loss": 17.5506,
+      "step": 11626
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.918168218248188e-05,
+      "loss": 17.6354,
+      "step": 11627
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.917803357073363e-05,
+      "loss": 14.8734,
+      "step": 11628
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.917438472336497e-05,
+      "loss": 15.8394,
+      "step": 11629
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.917073564040535e-05,
+      "loss": 15.4551,
+      "step": 11630
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.916708632188426e-05,
+      "loss": 17.0223,
+      "step": 11631
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.916343676783116e-05,
+      "loss": 17.1613,
+      "step": 11632
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.915978697827551e-05,
+      "loss": 14.3805,
+      "step": 11633
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.91561369532468e-05,
+      "loss": 16.937,
+      "step": 11634
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.915248669277448e-05,
+      "loss": 17.931,
+      "step": 11635
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.914883619688804e-05,
+      "loss": 15.2095,
+      "step": 11636
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.914518546561697e-05,
+      "loss": 15.9191,
+      "step": 11637
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.914153449899076e-05,
+      "loss": 14.1721,
+      "step": 11638
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.913788329703884e-05,
+      "loss": 18.3308,
+      "step": 11639
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.913423185979076e-05,
+      "loss": 15.5401,
+      "step": 11640
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.913058018727595e-05,
+      "loss": 16.8736,
+      "step": 11641
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.912692827952394e-05,
+      "loss": 16.8756,
+      "step": 11642
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.912327613656421e-05,
+      "loss": 17.4178,
+      "step": 11643
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.911962375842623e-05,
+      "loss": 15.9167,
+      "step": 11644
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.911597114513952e-05,
+      "loss": 17.8314,
+      "step": 11645
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.911231829673356e-05,
+      "loss": 15.2886,
+      "step": 11646
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.910866521323785e-05,
+      "loss": 16.0704,
+      "step": 11647
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.91050118946819e-05,
+      "loss": 16.9209,
+      "step": 11648
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.91013583410952e-05,
+      "loss": 16.2825,
+      "step": 11649
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.909770455250727e-05,
+      "loss": 17.0377,
+      "step": 11650
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.909405052894759e-05,
+      "loss": 16.8283,
+      "step": 11651
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.909039627044567e-05,
+      "loss": 17.2587,
+      "step": 11652
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.908674177703106e-05,
+      "loss": 15.5839,
+      "step": 11653
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.908308704873321e-05,
+      "loss": 16.9462,
+      "step": 11654
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.907943208558169e-05,
+      "loss": 17.0683,
+      "step": 11655
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.907577688760596e-05,
+      "loss": 14.4457,
+      "step": 11656
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.907212145483557e-05,
+      "loss": 15.6169,
+      "step": 11657
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.906846578730006e-05,
+      "loss": 14.8045,
+      "step": 11658
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.90648098850289e-05,
+      "loss": 14.0884,
+      "step": 11659
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.906115374805164e-05,
+      "loss": 15.4975,
+      "step": 11660
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.90574973763978e-05,
+      "loss": 15.8569,
+      "step": 11661
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.905384077009693e-05,
+      "loss": 16.1578,
+      "step": 11662
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.905018392917851e-05,
+      "loss": 16.4585,
+      "step": 11663
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.904652685367209e-05,
+      "loss": 14.2651,
+      "step": 11664
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.904286954360722e-05,
+      "loss": 16.8675,
+      "step": 11665
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.903921199901342e-05,
+      "loss": 17.3465,
+      "step": 11666
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.903555421992023e-05,
+      "loss": 14.6925,
+      "step": 11667
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.903189620635717e-05,
+      "loss": 17.0436,
+      "step": 11668
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.90282379583538e-05,
+      "loss": 16.1821,
+      "step": 11669
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.902457947593967e-05,
+      "loss": 15.7074,
+      "step": 11670
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.902092075914429e-05,
+      "loss": 17.0966,
+      "step": 11671
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.901726180799722e-05,
+      "loss": 15.6291,
+      "step": 11672
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.901360262252802e-05,
+      "loss": 15.4142,
+      "step": 11673
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.900994320276622e-05,
+      "loss": 14.6287,
+      "step": 11674
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.900628354874138e-05,
+      "loss": 14.1679,
+      "step": 11675
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.900262366048305e-05,
+      "loss": 15.51,
+      "step": 11676
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.899896353802079e-05,
+      "loss": 14.873,
+      "step": 11677
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.899530318138417e-05,
+      "loss": 17.3691,
+      "step": 11678
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.899164259060272e-05,
+      "loss": 15.6995,
+      "step": 11679
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.898798176570601e-05,
+      "loss": 14.9211,
+      "step": 11680
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.898432070672359e-05,
+      "loss": 15.9166,
+      "step": 11681
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.898065941368507e-05,
+      "loss": 18.5038,
+      "step": 11682
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.897699788661996e-05,
+      "loss": 17.3337,
+      "step": 11683
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.897333612555785e-05,
+      "loss": 14.2321,
+      "step": 11684
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.896967413052832e-05,
+      "loss": 15.5825,
+      "step": 11685
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.896601190156094e-05,
+      "loss": 15.2514,
+      "step": 11686
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.896234943868526e-05,
+      "loss": 14.742,
+      "step": 11687
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.895868674193089e-05,
+      "loss": 16.9257,
+      "step": 11688
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.895502381132737e-05,
+      "loss": 14.8627,
+      "step": 11689
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.895136064690432e-05,
+      "loss": 14.9802,
+      "step": 11690
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.894769724869128e-05,
+      "loss": 14.1602,
+      "step": 11691
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.894403361671784e-05,
+      "loss": 14.9346,
+      "step": 11692
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.894036975101363e-05,
+      "loss": 14.2975,
+      "step": 11693
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.89367056516082e-05,
+      "loss": 15.0654,
+      "step": 11694
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.893304131853111e-05,
+      "loss": 15.4667,
+      "step": 11695
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.892937675181203e-05,
+      "loss": 15.8514,
+      "step": 11696
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.892571195148046e-05,
+      "loss": 16.033,
+      "step": 11697
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.892204691756607e-05,
+      "loss": 14.8456,
+      "step": 11698
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.891838165009841e-05,
+      "loss": 15.9433,
+      "step": 11699
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.891471614910711e-05,
+      "loss": 18.1798,
+      "step": 11700
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.891105041462173e-05,
+      "loss": 16.2257,
+      "step": 11701
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.890738444667191e-05,
+      "loss": 14.7854,
+      "step": 11702
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.890371824528723e-05,
+      "loss": 17.954,
+      "step": 11703
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.89000518104973e-05,
+      "loss": 14.0348,
+      "step": 11704
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.889638514233175e-05,
+      "loss": 14.6007,
+      "step": 11705
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.889271824082016e-05,
+      "loss": 15.0423,
+      "step": 11706
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.888905110599216e-05,
+      "loss": 16.2845,
+      "step": 11707
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.888538373787734e-05,
+      "loss": 16.6669,
+      "step": 11708
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.888171613650534e-05,
+      "loss": 15.9832,
+      "step": 11709
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.887804830190578e-05,
+      "loss": 15.0484,
+      "step": 11710
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.887438023410824e-05,
+      "loss": 18.0191,
+      "step": 11711
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.88707119331424e-05,
+      "loss": 16.062,
+      "step": 11712
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.886704339903784e-05,
+      "loss": 14.4438,
+      "step": 11713
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.886337463182417e-05,
+      "loss": 15.4553,
+      "step": 11714
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.885970563153106e-05,
+      "loss": 16.2702,
+      "step": 11715
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.88560363981881e-05,
+      "loss": 13.3785,
+      "step": 11716
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.885236693182497e-05,
+      "loss": 16.0162,
+      "step": 11717
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.884869723247124e-05,
+      "loss": 16.7073,
+      "step": 11718
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.884502730015658e-05,
+      "loss": 15.6589,
+      "step": 11719
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.884135713491063e-05,
+      "loss": 16.0251,
+      "step": 11720
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.8837686736763e-05,
+      "loss": 15.7337,
+      "step": 11721
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.883401610574336e-05,
+      "loss": 16.5102,
+      "step": 11722
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.883034524188134e-05,
+      "loss": 14.7832,
+      "step": 11723
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.882667414520657e-05,
+      "loss": 15.3589,
+      "step": 11724
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.882300281574872e-05,
+      "loss": 17.3803,
+      "step": 11725
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.881933125353741e-05,
+      "loss": 17.808,
+      "step": 11726
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.881565945860229e-05,
+      "loss": 14.1713,
+      "step": 11727
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.881198743097306e-05,
+      "loss": 16.2593,
+      "step": 11728
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.880831517067932e-05,
+      "loss": 15.2924,
+      "step": 11729
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.880464267775073e-05,
+      "loss": 14.9351,
+      "step": 11730
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.880096995221695e-05,
+      "loss": 15.4477,
+      "step": 11731
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.879729699410766e-05,
+      "loss": 16.9067,
+      "step": 11732
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.87936238034525e-05,
+      "loss": 14.2377,
+      "step": 11733
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.878995038028114e-05,
+      "loss": 17.3289,
+      "step": 11734
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.878627672462324e-05,
+      "loss": 17.6359,
+      "step": 11735
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.878260283650846e-05,
+      "loss": 14.9117,
+      "step": 11736
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.877892871596647e-05,
+      "loss": 16.8768,
+      "step": 11737
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.877525436302696e-05,
+      "loss": 15.3586,
+      "step": 11738
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.877157977771956e-05,
+      "loss": 15.0682,
+      "step": 11739
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.876790496007397e-05,
+      "loss": 15.2143,
+      "step": 11740
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.87642299101199e-05,
+      "loss": 15.2588,
+      "step": 11741
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.876055462788695e-05,
+      "loss": 15.5157,
+      "step": 11742
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.875687911340486e-05,
+      "loss": 16.3354,
+      "step": 11743
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.875320336670328e-05,
+      "loss": 16.2558,
+      "step": 11744
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.874952738781189e-05,
+      "loss": 13.4592,
+      "step": 11745
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.874585117676041e-05,
+      "loss": 15.3442,
+      "step": 11746
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.87421747335785e-05,
+      "loss": 16.8064,
+      "step": 11747
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.873849805829584e-05,
+      "loss": 13.8101,
+      "step": 11748
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.873482115094213e-05,
+      "loss": 15.7462,
+      "step": 11749
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.873114401154709e-05,
+      "loss": 16.0927,
+      "step": 11750
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.872746664014035e-05,
+      "loss": 16.2699,
+      "step": 11751
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.872378903675167e-05,
+      "loss": 15.0951,
+      "step": 11752
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.872011120141072e-05,
+      "loss": 16.5416,
+      "step": 11753
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.871643313414718e-05,
+      "loss": 15.2485,
+      "step": 11754
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.871275483499079e-05,
+      "loss": 16.2959,
+      "step": 11755
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.870907630397123e-05,
+      "loss": 14.4121,
+      "step": 11756
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.870539754111822e-05,
+      "loss": 15.4628,
+      "step": 11757
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.870171854646144e-05,
+      "loss": 15.5428,
+      "step": 11758
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.869803932003061e-05,
+      "loss": 17.0084,
+      "step": 11759
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.869435986185547e-05,
+      "loss": 15.4636,
+      "step": 11760
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.86906801719657e-05,
+      "loss": 14.2305,
+      "step": 11761
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.868700025039102e-05,
+      "loss": 15.3843,
+      "step": 11762
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.868332009716114e-05,
+      "loss": 15.6457,
+      "step": 11763
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.86796397123058e-05,
+      "loss": 14.2838,
+      "step": 11764
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.867595909585473e-05,
+      "loss": 16.3952,
+      "step": 11765
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.867227824783761e-05,
+      "loss": 17.1018,
+      "step": 11766
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.866859716828418e-05,
+      "loss": 14.0985,
+      "step": 11767
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.866491585722417e-05,
+      "loss": 15.9654,
+      "step": 11768
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.866123431468732e-05,
+      "loss": 15.3473,
+      "step": 11769
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.865755254070333e-05,
+      "loss": 16.2765,
+      "step": 11770
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.865387053530196e-05,
+      "loss": 15.0483,
+      "step": 11771
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.865018829851295e-05,
+      "loss": 14.0387,
+      "step": 11772
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.864650583036598e-05,
+      "loss": 14.9124,
+      "step": 11773
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.864282313089084e-05,
+      "loss": 17.7375,
+      "step": 11774
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.863914020011724e-05,
+      "loss": 15.4114,
+      "step": 11775
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.863545703807496e-05,
+      "loss": 15.9582,
+      "step": 11776
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.863177364479368e-05,
+      "loss": 16.4311,
+      "step": 11777
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.86280900203032e-05,
+      "loss": 15.2504,
+      "step": 11778
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.862440616463324e-05,
+      "loss": 16.9438,
+      "step": 11779
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.862072207781357e-05,
+      "loss": 17.2538,
+      "step": 11780
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.86170377598739e-05,
+      "loss": 16.6761,
+      "step": 11781
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.861335321084402e-05,
+      "loss": 17.5272,
+      "step": 11782
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.860966843075368e-05,
+      "loss": 18.5112,
+      "step": 11783
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.860598341963262e-05,
+      "loss": 15.3467,
+      "step": 11784
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.860229817751059e-05,
+      "loss": 14.4062,
+      "step": 11785
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.859861270441735e-05,
+      "loss": 15.1552,
+      "step": 11786
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.859492700038269e-05,
+      "loss": 16.543,
+      "step": 11787
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.859124106543636e-05,
+      "loss": 15.7634,
+      "step": 11788
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.858755489960811e-05,
+      "loss": 16.1957,
+      "step": 11789
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.858386850292774e-05,
+      "loss": 16.2594,
+      "step": 11790
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.858018187542497e-05,
+      "loss": 17.2827,
+      "step": 11791
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.857649501712962e-05,
+      "loss": 14.2735,
+      "step": 11792
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.857280792807143e-05,
+      "loss": 14.8415,
+      "step": 11793
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.856912060828017e-05,
+      "loss": 18.8509,
+      "step": 11794
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.856543305778565e-05,
+      "loss": 15.7059,
+      "step": 11795
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.856174527661761e-05,
+      "loss": 15.9952,
+      "step": 11796
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.855805726480587e-05,
+      "loss": 15.9611,
+      "step": 11797
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.855436902238017e-05,
+      "loss": 15.3844,
+      "step": 11798
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.855068054937031e-05,
+      "loss": 15.9689,
+      "step": 11799
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.854699184580609e-05,
+      "loss": 15.2075,
+      "step": 11800
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.854330291171729e-05,
+      "loss": 17.1732,
+      "step": 11801
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.853961374713368e-05,
+      "loss": 15.5157,
+      "step": 11802
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.853592435208507e-05,
+      "loss": 18.0252,
+      "step": 11803
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.853223472660125e-05,
+      "loss": 17.1252,
+      "step": 11804
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.852854487071202e-05,
+      "loss": 16.3588,
+      "step": 11805
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.852485478444717e-05,
+      "loss": 16.2658,
+      "step": 11806
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.852116446783649e-05,
+      "loss": 13.6876,
+      "step": 11807
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.851747392090978e-05,
+      "loss": 15.2871,
+      "step": 11808
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.851378314369686e-05,
+      "loss": 16.1929,
+      "step": 11809
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.851009213622753e-05,
+      "loss": 16.1278,
+      "step": 11810
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.850640089853159e-05,
+      "loss": 15.9521,
+      "step": 11811
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.850270943063884e-05,
+      "loss": 15.3691,
+      "step": 11812
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.84990177325791e-05,
+      "loss": 15.7147,
+      "step": 11813
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.849532580438218e-05,
+      "loss": 15.005,
+      "step": 11814
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.84916336460779e-05,
+      "loss": 13.5992,
+      "step": 11815
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.848794125769605e-05,
+      "loss": 15.9281,
+      "step": 11816
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.848424863926649e-05,
+      "loss": 15.8963,
+      "step": 11817
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.848055579081899e-05,
+      "loss": 16.4336,
+      "step": 11818
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.847686271238341e-05,
+      "loss": 18.4488,
+      "step": 11819
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.847316940398954e-05,
+      "loss": 16.6514,
+      "step": 11820
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.846947586566723e-05,
+      "loss": 15.5671,
+      "step": 11821
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.846578209744631e-05,
+      "loss": 13.3284,
+      "step": 11822
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.846208809935658e-05,
+      "loss": 15.5541,
+      "step": 11823
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.845839387142788e-05,
+      "loss": 16.0609,
+      "step": 11824
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.845469941369005e-05,
+      "loss": 16.7388,
+      "step": 11825
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.845100472617296e-05,
+      "loss": 17.0787,
+      "step": 11826
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.844730980890636e-05,
+      "loss": 16.1937,
+      "step": 11827
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.844361466192015e-05,
+      "loss": 14.7766,
+      "step": 11828
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.843991928524415e-05,
+      "loss": 16.0886,
+      "step": 11829
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.843622367890822e-05,
+      "loss": 16.4433,
+      "step": 11830
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.843252784294218e-05,
+      "loss": 15.8523,
+      "step": 11831
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.842883177737586e-05,
+      "loss": 16.4235,
+      "step": 11832
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.842513548223915e-05,
+      "loss": 15.4466,
+      "step": 11833
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.842143895756188e-05,
+      "loss": 14.7863,
+      "step": 11834
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.84177422033739e-05,
+      "loss": 15.5578,
+      "step": 11835
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.841404521970505e-05,
+      "loss": 15.207,
+      "step": 11836
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.841034800658521e-05,
+      "loss": 16.1337,
+      "step": 11837
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.840665056404421e-05,
+      "loss": 15.2281,
+      "step": 11838
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.840295289211192e-05,
+      "loss": 15.1819,
+      "step": 11839
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.83992549908182e-05,
+      "loss": 16.0714,
+      "step": 11840
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.839555686019291e-05,
+      "loss": 15.3923,
+      "step": 11841
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.839185850026591e-05,
+      "loss": 15.6749,
+      "step": 11842
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.838815991106708e-05,
+      "loss": 15.6819,
+      "step": 11843
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.838446109262627e-05,
+      "loss": 16.7998,
+      "step": 11844
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.838076204497334e-05,
+      "loss": 18.7608,
+      "step": 11845
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.837706276813819e-05,
+      "loss": 17.5332,
+      "step": 11846
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.837336326215068e-05,
+      "loss": 16.0437,
+      "step": 11847
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.836966352704067e-05,
+      "loss": 16.3646,
+      "step": 11848
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.836596356283805e-05,
+      "loss": 15.9491,
+      "step": 11849
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.836226336957271e-05,
+      "loss": 16.2317,
+      "step": 11850
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.83585629472745e-05,
+      "loss": 17.6386,
+      "step": 11851
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.835486229597334e-05,
+      "loss": 15.4523,
+      "step": 11852
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.83511614156991e-05,
+      "loss": 17.3818,
+      "step": 11853
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.834746030648163e-05,
+      "loss": 14.4188,
+      "step": 11854
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.834375896835086e-05,
+      "loss": 15.8669,
+      "step": 11855
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.834005740133665e-05,
+      "loss": 15.8134,
+      "step": 11856
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.833635560546893e-05,
+      "loss": 14.0306,
+      "step": 11857
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.833265358077755e-05,
+      "loss": 14.5424,
+      "step": 11858
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.832895132729243e-05,
+      "loss": 16.1601,
+      "step": 11859
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.832524884504347e-05,
+      "loss": 15.4838,
+      "step": 11860
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.832154613406055e-05,
+      "loss": 15.4196,
+      "step": 11861
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.831784319437359e-05,
+      "loss": 15.1776,
+      "step": 11862
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.831414002601247e-05,
+      "loss": 17.5889,
+      "step": 11863
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.831043662900711e-05,
+      "loss": 17.2384,
+      "step": 11864
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.830673300338742e-05,
+      "loss": 15.3925,
+      "step": 11865
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.830302914918328e-05,
+      "loss": 15.2381,
+      "step": 11866
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.829932506642463e-05,
+      "loss": 17.1181,
+      "step": 11867
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.829562075514138e-05,
+      "loss": 15.9905,
+      "step": 11868
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.829191621536342e-05,
+      "loss": 15.8995,
+      "step": 11869
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.828821144712068e-05,
+      "loss": 14.085,
+      "step": 11870
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.828450645044309e-05,
+      "loss": 16.5996,
+      "step": 11871
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.828080122536054e-05,
+      "loss": 14.8356,
+      "step": 11872
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.827709577190298e-05,
+      "loss": 15.9316,
+      "step": 11873
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.82733900901003e-05,
+      "loss": 17.0239,
+      "step": 11874
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.826968417998246e-05,
+      "loss": 14.3714,
+      "step": 11875
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.826597804157936e-05,
+      "loss": 15.2571,
+      "step": 11876
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.826227167492093e-05,
+      "loss": 16.3802,
+      "step": 11877
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.825856508003713e-05,
+      "loss": 14.8917,
+      "step": 11878
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.825485825695784e-05,
+      "loss": 16.3994,
+      "step": 11879
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.825115120571305e-05,
+      "loss": 13.8871,
+      "step": 11880
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.824744392633264e-05,
+      "loss": 15.4925,
+      "step": 11881
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.824373641884659e-05,
+      "loss": 15.7676,
+      "step": 11882
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.824002868328483e-05,
+      "loss": 15.4437,
+      "step": 11883
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.823632071967728e-05,
+      "loss": 15.622,
+      "step": 11884
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.823261252805389e-05,
+      "loss": 16.795,
+      "step": 11885
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.822890410844463e-05,
+      "loss": 15.9257,
+      "step": 11886
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.822519546087943e-05,
+      "loss": 15.7204,
+      "step": 11887
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.822148658538824e-05,
+      "loss": 17.7722,
+      "step": 11888
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.8217777482001e-05,
+      "loss": 15.5502,
+      "step": 11889
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.821406815074768e-05,
+      "loss": 17.32,
+      "step": 11890
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.82103585916582e-05,
+      "loss": 16.6231,
+      "step": 11891
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.820664880476256e-05,
+      "loss": 15.3079,
+      "step": 11892
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.820293879009069e-05,
+      "loss": 15.6245,
+      "step": 11893
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.819922854767254e-05,
+      "loss": 14.7138,
+      "step": 11894
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.819551807753811e-05,
+      "loss": 17.4321,
+      "step": 11895
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.819180737971732e-05,
+      "loss": 15.277,
+      "step": 11896
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.818809645424016e-05,
+      "loss": 17.0956,
+      "step": 11897
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.818438530113661e-05,
+      "loss": 16.5208,
+      "step": 11898
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.818067392043661e-05,
+      "loss": 15.5102,
+      "step": 11899
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.817696231217014e-05,
+      "loss": 16.5509,
+      "step": 11900
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.817325047636717e-05,
+      "loss": 17.2454,
+      "step": 11901
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.816953841305767e-05,
+      "loss": 14.3172,
+      "step": 11902
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.816582612227163e-05,
+      "loss": 15.798,
+      "step": 11903
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.816211360403902e-05,
+      "loss": 15.5672,
+      "step": 11904
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.815840085838982e-05,
+      "loss": 16.6032,
+      "step": 11905
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.815468788535402e-05,
+      "loss": 16.335,
+      "step": 11906
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.815097468496158e-05,
+      "loss": 15.404,
+      "step": 11907
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.81472612572425e-05,
+      "loss": 17.6187,
+      "step": 11908
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.814354760222677e-05,
+      "loss": 15.6229,
+      "step": 11909
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.813983371994439e-05,
+      "loss": 15.9651,
+      "step": 11910
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.813611961042532e-05,
+      "loss": 15.2312,
+      "step": 11911
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.813240527369959e-05,
+      "loss": 16.4051,
+      "step": 11912
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.812869070979713e-05,
+      "loss": 15.2123,
+      "step": 11913
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.812497591874801e-05,
+      "loss": 15.8501,
+      "step": 11914
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.812126090058219e-05,
+      "loss": 14.2771,
+      "step": 11915
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.811754565532969e-05,
+      "loss": 16.9531,
+      "step": 11916
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.811383018302048e-05,
+      "loss": 16.3976,
+      "step": 11917
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.811011448368459e-05,
+      "loss": 14.0693,
+      "step": 11918
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.8106398557352e-05,
+      "loss": 14.7141,
+      "step": 11919
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.810268240405276e-05,
+      "loss": 17.5061,
+      "step": 11920
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.809896602381683e-05,
+      "loss": 15.7206,
+      "step": 11921
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.809524941667425e-05,
+      "loss": 15.6069,
+      "step": 11922
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.809153258265504e-05,
+      "loss": 16.2935,
+      "step": 11923
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.808781552178919e-05,
+      "loss": 15.7736,
+      "step": 11924
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.808409823410671e-05,
+      "loss": 16.7246,
+      "step": 11925
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.808038071963767e-05,
+      "loss": 16.166,
+      "step": 11926
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.807666297841203e-05,
+      "loss": 17.5997,
+      "step": 11927
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.807294501045985e-05,
+      "loss": 16.4095,
+      "step": 11928
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.806922681581113e-05,
+      "loss": 17.5543,
+      "step": 11929
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.806550839449591e-05,
+      "loss": 14.3904,
+      "step": 11930
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.806178974654421e-05,
+      "loss": 15.5214,
+      "step": 11931
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.805807087198607e-05,
+      "loss": 14.6838,
+      "step": 11932
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.80543517708515e-05,
+      "loss": 15.5,
+      "step": 11933
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.805063244317055e-05,
+      "loss": 15.3768,
+      "step": 11934
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.804691288897325e-05,
+      "loss": 17.5896,
+      "step": 11935
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.804319310828965e-05,
+      "loss": 16.458,
+      "step": 11936
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.803947310114976e-05,
+      "loss": 15.4428,
+      "step": 11937
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.803575286758364e-05,
+      "loss": 13.0233,
+      "step": 11938
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.803203240762132e-05,
+      "loss": 17.8651,
+      "step": 11939
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.802831172129285e-05,
+      "loss": 15.4542,
+      "step": 11940
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.80245908086283e-05,
+      "loss": 15.3477,
+      "step": 11941
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.802086966965767e-05,
+      "loss": 15.9177,
+      "step": 11942
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.801714830441105e-05,
+      "loss": 15.8698,
+      "step": 11943
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.801342671291844e-05,
+      "loss": 15.1576,
+      "step": 11944
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.800970489520996e-05,
+      "loss": 17.1954,
+      "step": 11945
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.800598285131562e-05,
+      "loss": 13.5388,
+      "step": 11946
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.800226058126547e-05,
+      "loss": 15.7749,
+      "step": 11947
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.799853808508961e-05,
+      "loss": 17.78,
+      "step": 11948
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.799481536281807e-05,
+      "loss": 17.3041,
+      "step": 11949
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.799109241448091e-05,
+      "loss": 16.2957,
+      "step": 11950
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.798736924010821e-05,
+      "loss": 16.3638,
+      "step": 11951
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.798364583973003e-05,
+      "loss": 15.3094,
+      "step": 11952
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.797992221337642e-05,
+      "loss": 15.0178,
+      "step": 11953
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.797619836107748e-05,
+      "loss": 16.5036,
+      "step": 11954
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.797247428286326e-05,
+      "loss": 15.3077,
+      "step": 11955
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.796874997876385e-05,
+      "loss": 17.0858,
+      "step": 11956
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.79650254488093e-05,
+      "loss": 16.0145,
+      "step": 11957
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.796130069302969e-05,
+      "loss": 16.8987,
+      "step": 11958
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.795757571145512e-05,
+      "loss": 16.2774,
+      "step": 11959
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.795385050411565e-05,
+      "loss": 17.5589,
+      "step": 11960
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.795012507104138e-05,
+      "loss": 17.1872,
+      "step": 11961
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.794639941226238e-05,
+      "loss": 16.5014,
+      "step": 11962
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.794267352780874e-05,
+      "loss": 17.6859,
+      "step": 11963
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.793894741771055e-05,
+      "loss": 15.35,
+      "step": 11964
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.793522108199789e-05,
+      "loss": 15.1249,
+      "step": 11965
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.793149452070085e-05,
+      "loss": 14.5868,
+      "step": 11966
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.792776773384956e-05,
+      "loss": 15.6794,
+      "step": 11967
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.792404072147405e-05,
+      "loss": 16.7727,
+      "step": 11968
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.792031348360448e-05,
+      "loss": 15.3058,
+      "step": 11969
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.791658602027092e-05,
+      "loss": 15.0434,
+      "step": 11970
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.791285833150344e-05,
+      "loss": 15.1369,
+      "step": 11971
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.790913041733221e-05,
+      "loss": 16.1817,
+      "step": 11972
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.790540227778726e-05,
+      "loss": 15.4994,
+      "step": 11973
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.790167391289876e-05,
+      "loss": 16.6599,
+      "step": 11974
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.789794532269678e-05,
+      "loss": 15.334,
+      "step": 11975
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.789421650721145e-05,
+      "loss": 15.5454,
+      "step": 11976
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.789048746647286e-05,
+      "loss": 15.8509,
+      "step": 11977
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.788675820051113e-05,
+      "loss": 15.8202,
+      "step": 11978
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.78830287093564e-05,
+      "loss": 16.6532,
+      "step": 11979
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.787929899303875e-05,
+      "loss": 14.9926,
+      "step": 11980
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.78755690515883e-05,
+      "loss": 16.2281,
+      "step": 11981
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.78718388850352e-05,
+      "loss": 14.6354,
+      "step": 11982
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.786810849340955e-05,
+      "loss": 14.4313,
+      "step": 11983
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.786437787674149e-05,
+      "loss": 13.136,
+      "step": 11984
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.786064703506113e-05,
+      "loss": 14.2457,
+      "step": 11985
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.785691596839862e-05,
+      "loss": 14.9032,
+      "step": 11986
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.785318467678404e-05,
+      "loss": 15.2726,
+      "step": 11987
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.784945316024756e-05,
+      "loss": 14.8851,
+      "step": 11988
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.784572141881932e-05,
+      "loss": 16.7733,
+      "step": 11989
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.784198945252945e-05,
+      "loss": 14.7468,
+      "step": 11990
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.783825726140805e-05,
+      "loss": 16.5879,
+      "step": 11991
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.78345248454853e-05,
+      "loss": 15.0251,
+      "step": 11992
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.783079220479133e-05,
+      "loss": 15.4074,
+      "step": 11993
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.782705933935628e-05,
+      "loss": 14.5355,
+      "step": 11994
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.782332624921027e-05,
+      "loss": 16.9849,
+      "step": 11995
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.78195929343835e-05,
+      "loss": 16.2872,
+      "step": 11996
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.781585939490607e-05,
+      "loss": 15.7894,
+      "step": 11997
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.781212563080816e-05,
+      "loss": 16.4388,
+      "step": 11998
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.780839164211988e-05,
+      "loss": 14.764,
+      "step": 11999
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.780465742887142e-05,
+      "loss": 14.763,
+      "step": 12000
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.780092299109293e-05,
+      "loss": 16.0971,
+      "step": 12001
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.779718832881455e-05,
+      "loss": 15.8769,
+      "step": 12002
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.779345344206647e-05,
+      "loss": 16.5136,
+      "step": 12003
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.778971833087881e-05,
+      "loss": 16.1236,
+      "step": 12004
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.778598299528176e-05,
+      "loss": 17.2342,
+      "step": 12005
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.778224743530546e-05,
+      "loss": 14.7425,
+      "step": 12006
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.777851165098012e-05,
+      "loss": 15.4168,
+      "step": 12007
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.777477564233586e-05,
+      "loss": 14.8951,
+      "step": 12008
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.777103940940286e-05,
+      "loss": 16.9682,
+      "step": 12009
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.776730295221132e-05,
+      "loss": 15.2122,
+      "step": 12010
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.776356627079137e-05,
+      "loss": 15.8799,
+      "step": 12011
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.775982936517322e-05,
+      "loss": 16.3366,
+      "step": 12012
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.775609223538702e-05,
+      "loss": 14.3731,
+      "step": 12013
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.775235488146297e-05,
+      "loss": 15.9072,
+      "step": 12014
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.774861730343125e-05,
+      "loss": 14.5496,
+      "step": 12015
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.774487950132202e-05,
+      "loss": 16.486,
+      "step": 12016
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.774114147516549e-05,
+      "loss": 17.0138,
+      "step": 12017
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.773740322499181e-05,
+      "loss": 13.6315,
+      "step": 12018
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.773366475083121e-05,
+      "loss": 17.5295,
+      "step": 12019
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.772992605271386e-05,
+      "loss": 13.2008,
+      "step": 12020
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.772618713066992e-05,
+      "loss": 15.7074,
+      "step": 12021
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.772244798472965e-05,
+      "loss": 15.0813,
+      "step": 12022
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.771870861492319e-05,
+      "loss": 13.362,
+      "step": 12023
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.771496902128075e-05,
+      "loss": 15.2896,
+      "step": 12024
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.771122920383254e-05,
+      "loss": 17.0905,
+      "step": 12025
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.770748916260875e-05,
+      "loss": 14.421,
+      "step": 12026
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.770374889763958e-05,
+      "loss": 15.2158,
+      "step": 12027
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.770000840895523e-05,
+      "loss": 16.3049,
+      "step": 12028
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.769626769658594e-05,
+      "loss": 15.7964,
+      "step": 12029
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.769252676056187e-05,
+      "loss": 16.0463,
+      "step": 12030
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.768878560091325e-05,
+      "loss": 15.5575,
+      "step": 12031
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.768504421767028e-05,
+      "loss": 13.6295,
+      "step": 12032
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.768130261086321e-05,
+      "loss": 15.3662,
+      "step": 12033
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.76775607805222e-05,
+      "loss": 14.8047,
+      "step": 12034
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.767381872667752e-05,
+      "loss": 14.1967,
+      "step": 12035
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.767007644935934e-05,
+      "loss": 16.8072,
+      "step": 12036
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.766633394859791e-05,
+      "loss": 15.7615,
+      "step": 12037
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.766259122442344e-05,
+      "loss": 16.5802,
+      "step": 12038
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.765884827686616e-05,
+      "loss": 15.7785,
+      "step": 12039
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.765510510595629e-05,
+      "loss": 17.6324,
+      "step": 12040
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.765136171172406e-05,
+      "loss": 14.984,
+      "step": 12041
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.76476180941997e-05,
+      "loss": 16.3538,
+      "step": 12042
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.764387425341343e-05,
+      "loss": 14.833,
+      "step": 12043
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.76401301893955e-05,
+      "loss": 16.3543,
+      "step": 12044
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.763638590217613e-05,
+      "loss": 17.5236,
+      "step": 12045
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.763264139178557e-05,
+      "loss": 16.4026,
+      "step": 12046
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.762889665825403e-05,
+      "loss": 13.878,
+      "step": 12047
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.762515170161179e-05,
+      "loss": 14.826,
+      "step": 12048
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.762140652188906e-05,
+      "loss": 17.4989,
+      "step": 12049
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.76176611191161e-05,
+      "loss": 15.8867,
+      "step": 12050
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.761391549332315e-05,
+      "loss": 16.0628,
+      "step": 12051
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.761016964454045e-05,
+      "loss": 17.0703,
+      "step": 12052
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.760642357279826e-05,
+      "loss": 16.001,
+      "step": 12053
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.760267727812682e-05,
+      "loss": 12.948,
+      "step": 12054
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.75989307605564e-05,
+      "loss": 15.2222,
+      "step": 12055
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.759518402011722e-05,
+      "loss": 15.3836,
+      "step": 12056
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.759143705683958e-05,
+      "loss": 14.1705,
+      "step": 12057
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.75876898707537e-05,
+      "loss": 15.1579,
+      "step": 12058
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.758394246188987e-05,
+      "loss": 15.8592,
+      "step": 12059
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.758019483027832e-05,
+      "loss": 15.084,
+      "step": 12060
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.757644697594934e-05,
+      "loss": 14.3743,
+      "step": 12061
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.757269889893319e-05,
+      "loss": 16.4194,
+      "step": 12062
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.756895059926013e-05,
+      "loss": 17.4992,
+      "step": 12063
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.756520207696041e-05,
+      "loss": 16.4455,
+      "step": 12064
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.756145333206432e-05,
+      "loss": 15.4995,
+      "step": 12065
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.755770436460216e-05,
+      "loss": 16.781,
+      "step": 12066
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.755395517460416e-05,
+      "loss": 15.2487,
+      "step": 12067
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.755020576210061e-05,
+      "loss": 15.1798,
+      "step": 12068
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.75464561271218e-05,
+      "loss": 16.9243,
+      "step": 12069
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.754270626969797e-05,
+      "loss": 15.8269,
+      "step": 12070
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.753895618985947e-05,
+      "loss": 15.3418,
+      "step": 12071
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.753520588763652e-05,
+      "loss": 16.4786,
+      "step": 12072
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.753145536305943e-05,
+      "loss": 17.2778,
+      "step": 12073
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.752770461615848e-05,
+      "loss": 17.0641,
+      "step": 12074
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.752395364696398e-05,
+      "loss": 16.4864,
+      "step": 12075
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.752020245550618e-05,
+      "loss": 15.0002,
+      "step": 12076
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.75164510418154e-05,
+      "loss": 16.3316,
+      "step": 12077
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.751269940592194e-05,
+      "loss": 16.7268,
+      "step": 12078
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.750894754785607e-05,
+      "loss": 14.7478,
+      "step": 12079
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.75051954676481e-05,
+      "loss": 15.256,
+      "step": 12080
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.750144316532833e-05,
+      "loss": 13.8075,
+      "step": 12081
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.749769064092706e-05,
+      "loss": 16.0438,
+      "step": 12082
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.749393789447459e-05,
+      "loss": 17.2192,
+      "step": 12083
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.749018492600125e-05,
+      "loss": 16.5138,
+      "step": 12084
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.748643173553729e-05,
+      "loss": 16.8839,
+      "step": 12085
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.748267832311307e-05,
+      "loss": 15.2878,
+      "step": 12086
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.747892468875888e-05,
+      "loss": 18.0273,
+      "step": 12087
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.747517083250504e-05,
+      "loss": 15.6865,
+      "step": 12088
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.747141675438184e-05,
+      "loss": 16.9864,
+      "step": 12089
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.746766245441963e-05,
+      "loss": 15.3739,
+      "step": 12090
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.746390793264869e-05,
+      "loss": 14.3927,
+      "step": 12091
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.746015318909937e-05,
+      "loss": 15.2923,
+      "step": 12092
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.745639822380197e-05,
+      "loss": 16.2682,
+      "step": 12093
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.745264303678683e-05,
+      "loss": 17.813,
+      "step": 12094
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.744888762808426e-05,
+      "loss": 15.9788,
+      "step": 12095
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.744513199772459e-05,
+      "loss": 16.744,
+      "step": 12096
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.744137614573814e-05,
+      "loss": 16.19,
+      "step": 12097
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.743762007215525e-05,
+      "loss": 13.7258,
+      "step": 12098
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.743386377700625e-05,
+      "loss": 15.8043,
+      "step": 12099
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.743010726032148e-05,
+      "loss": 14.3296,
+      "step": 12100
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.742635052213126e-05,
+      "loss": 15.5212,
+      "step": 12101
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.742259356246593e-05,
+      "loss": 14.824,
+      "step": 12102
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.741883638135585e-05,
+      "loss": 16.9192,
+      "step": 12103
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.741507897883132e-05,
+      "loss": 15.6895,
+      "step": 12104
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.741132135492271e-05,
+      "loss": 17.0084,
+      "step": 12105
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.740756350966036e-05,
+      "loss": 17.3957,
+      "step": 12106
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.740380544307461e-05,
+      "loss": 16.6953,
+      "step": 12107
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.740004715519583e-05,
+      "loss": 17.2114,
+      "step": 12108
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.739628864605434e-05,
+      "loss": 15.6983,
+      "step": 12109
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.739252991568049e-05,
+      "loss": 15.1783,
+      "step": 12110
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.738877096410465e-05,
+      "loss": 15.4753,
+      "step": 12111
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.738501179135718e-05,
+      "loss": 16.5181,
+      "step": 12112
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.73812523974684e-05,
+      "loss": 15.8454,
+      "step": 12113
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.737749278246871e-05,
+      "loss": 16.5562,
+      "step": 12114
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.737373294638846e-05,
+      "loss": 15.8917,
+      "step": 12115
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.736997288925799e-05,
+      "loss": 17.4687,
+      "step": 12116
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.736621261110768e-05,
+      "loss": 15.7473,
+      "step": 12117
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.73624521119679e-05,
+      "loss": 16.2695,
+      "step": 12118
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.7358691391869e-05,
+      "loss": 18.5859,
+      "step": 12119
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.735493045084137e-05,
+      "loss": 16.754,
+      "step": 12120
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.735116928891536e-05,
+      "loss": 15.0033,
+      "step": 12121
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.734740790612136e-05,
+      "loss": 15.4823,
+      "step": 12122
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.734364630248972e-05,
+      "loss": 16.8053,
+      "step": 12123
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.733988447805084e-05,
+      "loss": 15.6199,
+      "step": 12124
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.73361224328351e-05,
+      "loss": 16.7539,
+      "step": 12125
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.733236016687286e-05,
+      "loss": 16.7232,
+      "step": 12126
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.732859768019451e-05,
+      "loss": 14.4964,
+      "step": 12127
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.732483497283044e-05,
+      "loss": 14.6265,
+      "step": 12128
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.732107204481102e-05,
+      "loss": 16.6135,
+      "step": 12129
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.731730889616665e-05,
+      "loss": 14.6414,
+      "step": 12130
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.73135455269277e-05,
+      "loss": 15.4951,
+      "step": 12131
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.730978193712462e-05,
+      "loss": 14.181,
+      "step": 12132
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.73060181267877e-05,
+      "loss": 15.115,
+      "step": 12133
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.730225409594742e-05,
+      "loss": 15.4809,
+      "step": 12134
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.729848984463415e-05,
+      "loss": 16.428,
+      "step": 12135
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.729472537287827e-05,
+      "loss": 15.7831,
+      "step": 12136
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.72909606807102e-05,
+      "loss": 17.3352,
+      "step": 12137
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.728719576816032e-05,
+      "loss": 14.4352,
+      "step": 12138
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.728343063525905e-05,
+      "loss": 14.0309,
+      "step": 12139
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.727966528203678e-05,
+      "loss": 14.3624,
+      "step": 12140
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.727589970852393e-05,
+      "loss": 14.9994,
+      "step": 12141
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.727213391475093e-05,
+      "loss": 16.503,
+      "step": 12142
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.726836790074813e-05,
+      "loss": 13.881,
+      "step": 12143
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.726460166654598e-05,
+      "loss": 17.4837,
+      "step": 12144
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.726083521217491e-05,
+      "loss": 16.2684,
+      "step": 12145
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.72570685376653e-05,
+      "loss": 17.187,
+      "step": 12146
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.725330164304758e-05,
+      "loss": 17.2551,
+      "step": 12147
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.724953452835217e-05,
+      "loss": 17.147,
+      "step": 12148
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.72457671936095e-05,
+      "loss": 14.7093,
+      "step": 12149
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.724199963884997e-05,
+      "loss": 14.5785,
+      "step": 12150
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.723823186410402e-05,
+      "loss": 15.3898,
+      "step": 12151
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.723446386940208e-05,
+      "loss": 16.4043,
+      "step": 12152
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.723069565477455e-05,
+      "loss": 16.3399,
+      "step": 12153
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.722692722025188e-05,
+      "loss": 18.7005,
+      "step": 12154
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.722315856586453e-05,
+      "loss": 15.3325,
+      "step": 12155
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.721938969164287e-05,
+      "loss": 17.3511,
+      "step": 12156
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.721562059761739e-05,
+      "loss": 15.4196,
+      "step": 12157
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.721185128381849e-05,
+      "loss": 14.9552,
+      "step": 12158
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.720808175027663e-05,
+      "loss": 15.7642,
+      "step": 12159
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.720431199702224e-05,
+      "loss": 14.9864,
+      "step": 12160
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.720054202408575e-05,
+      "loss": 17.6323,
+      "step": 12161
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.719677183149764e-05,
+      "loss": 16.6246,
+      "step": 12162
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.719300141928832e-05,
+      "loss": 15.6808,
+      "step": 12163
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.718923078748825e-05,
+      "loss": 14.5107,
+      "step": 12164
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.718545993612788e-05,
+      "loss": 15.8663,
+      "step": 12165
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.718168886523766e-05,
+      "loss": 17.6703,
+      "step": 12166
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.717791757484804e-05,
+      "loss": 16.2883,
+      "step": 12167
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.717414606498947e-05,
+      "loss": 15.6604,
+      "step": 12168
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.717037433569242e-05,
+      "loss": 15.8765,
+      "step": 12169
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.716660238698734e-05,
+      "loss": 16.799,
+      "step": 12170
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.71628302189047e-05,
+      "loss": 15.5422,
+      "step": 12171
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.715905783147492e-05,
+      "loss": 16.4038,
+      "step": 12172
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.715528522472851e-05,
+      "loss": 16.219,
+      "step": 12173
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.715151239869591e-05,
+      "loss": 15.0711,
+      "step": 12174
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.714773935340761e-05,
+      "loss": 17.1151,
+      "step": 12175
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.714396608889405e-05,
+      "loss": 16.5541,
+      "step": 12176
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.714019260518573e-05,
+      "loss": 16.0351,
+      "step": 12177
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.71364189023131e-05,
+      "loss": 16.101,
+      "step": 12178
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.713264498030661e-05,
+      "loss": 16.1871,
+      "step": 12179
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.71288708391968e-05,
+      "loss": 16.5074,
+      "step": 12180
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.712509647901409e-05,
+      "loss": 14.3105,
+      "step": 12181
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.712132189978899e-05,
+      "loss": 16.0588,
+      "step": 12182
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.711754710155197e-05,
+      "loss": 15.8156,
+      "step": 12183
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.711377208433351e-05,
+      "loss": 16.4694,
+      "step": 12184
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.710999684816411e-05,
+      "loss": 16.5275,
+      "step": 12185
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.710622139307423e-05,
+      "loss": 15.2226,
+      "step": 12186
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.710244571909437e-05,
+      "loss": 14.375,
+      "step": 12187
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.709866982625504e-05,
+      "loss": 16.3745,
+      "step": 12188
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.709489371458669e-05,
+      "loss": 17.759,
+      "step": 12189
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.709111738411985e-05,
+      "loss": 16.0408,
+      "step": 12190
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.7087340834885e-05,
+      "loss": 19.3021,
+      "step": 12191
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.708356406691262e-05,
+      "loss": 15.93,
+      "step": 12192
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 7.707978708023325e-05,
+      "loss": 16.0729,
+      "step": 12193
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.707600987487736e-05,
+      "loss": 16.7541,
+      "step": 12194
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.707223245087545e-05,
+      "loss": 14.3271,
+      "step": 12195
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.706845480825803e-05,
+      "loss": 17.3386,
+      "step": 12196
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.706467694705561e-05,
+      "loss": 16.16,
+      "step": 12197
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.706089886729869e-05,
+      "loss": 16.377,
+      "step": 12198
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.705712056901779e-05,
+      "loss": 15.4262,
+      "step": 12199
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.705334205224342e-05,
+      "loss": 14.2302,
+      "step": 12200
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.704956331700608e-05,
+      "loss": 15.2583,
+      "step": 12201
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.704578436333629e-05,
+      "loss": 14.7252,
+      "step": 12202
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.704200519126457e-05,
+      "loss": 17.2787,
+      "step": 12203
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.703822580082143e-05,
+      "loss": 16.5251,
+      "step": 12204
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.703444619203739e-05,
+      "loss": 17.4022,
+      "step": 12205
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.703066636494299e-05,
+      "loss": 16.0369,
+      "step": 12206
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.702688631956871e-05,
+      "loss": 14.9593,
+      "step": 12207
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.702310605594512e-05,
+      "loss": 15.5633,
+      "step": 12208
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.701932557410274e-05,
+      "loss": 15.0527,
+      "step": 12209
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.701554487407207e-05,
+      "loss": 18.1592,
+      "step": 12210
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.701176395588367e-05,
+      "loss": 15.6797,
+      "step": 12211
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.700798281956807e-05,
+      "loss": 15.4337,
+      "step": 12212
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.700420146515579e-05,
+      "loss": 15.3595,
+      "step": 12213
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.700041989267736e-05,
+      "loss": 14.9105,
+      "step": 12214
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.699663810216333e-05,
+      "loss": 15.88,
+      "step": 12215
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.699285609364424e-05,
+      "loss": 17.2122,
+      "step": 12216
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.698907386715061e-05,
+      "loss": 17.2994,
+      "step": 12217
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.698529142271302e-05,
+      "loss": 15.9793,
+      "step": 12218
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.698150876036199e-05,
+      "loss": 15.6638,
+      "step": 12219
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.697772588012808e-05,
+      "loss": 17.0617,
+      "step": 12220
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.69739427820418e-05,
+      "loss": 15.4865,
+      "step": 12221
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.697015946613376e-05,
+      "loss": 14.9163,
+      "step": 12222
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.696637593243444e-05,
+      "loss": 15.9638,
+      "step": 12223
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.696259218097446e-05,
+      "loss": 16.1909,
+      "step": 12224
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.695880821178433e-05,
+      "loss": 17.2153,
+      "step": 12225
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.695502402489463e-05,
+      "loss": 15.2927,
+      "step": 12226
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.69512396203359e-05,
+      "loss": 15.5876,
+      "step": 12227
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.694745499813874e-05,
+      "loss": 16.8655,
+      "step": 12228
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.694367015833365e-05,
+      "loss": 15.4153,
+      "step": 12229
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.693988510095125e-05,
+      "loss": 17.0282,
+      "step": 12230
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.693609982602205e-05,
+      "loss": 14.8706,
+      "step": 12231
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.693231433357668e-05,
+      "loss": 15.7728,
+      "step": 12232
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.692852862364566e-05,
+      "loss": 16.8276,
+      "step": 12233
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.69247426962596e-05,
+      "loss": 14.6176,
+      "step": 12234
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.692095655144901e-05,
+      "loss": 13.9797,
+      "step": 12235
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.691717018924454e-05,
+      "loss": 15.4274,
+      "step": 12236
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.691338360967673e-05,
+      "loss": 15.4295,
+      "step": 12237
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.690959681277614e-05,
+      "loss": 17.1014,
+      "step": 12238
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.690580979857338e-05,
+      "loss": 15.9958,
+      "step": 12239
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.6902022567099e-05,
+      "loss": 16.9269,
+      "step": 12240
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.689823511838363e-05,
+      "loss": 16.8383,
+      "step": 12241
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.689444745245782e-05,
+      "loss": 17.3583,
+      "step": 12242
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.689065956935217e-05,
+      "loss": 17.9152,
+      "step": 12243
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.688687146909724e-05,
+      "loss": 13.7054,
+      "step": 12244
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.688308315172364e-05,
+      "loss": 17.5276,
+      "step": 12245
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.687929461726198e-05,
+      "loss": 15.7087,
+      "step": 12246
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.687550586574282e-05,
+      "loss": 15.5507,
+      "step": 12247
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.687171689719677e-05,
+      "loss": 15.4606,
+      "step": 12248
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.686792771165445e-05,
+      "loss": 15.4438,
+      "step": 12249
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.686413830914642e-05,
+      "loss": 18.6413,
+      "step": 12250
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.68603486897033e-05,
+      "loss": 15.3836,
+      "step": 12251
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.685655885335569e-05,
+      "loss": 16.916,
+      "step": 12252
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.685276880013419e-05,
+      "loss": 14.3602,
+      "step": 12253
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.68489785300694e-05,
+      "loss": 15.6709,
+      "step": 12254
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.684518804319195e-05,
+      "loss": 15.2857,
+      "step": 12255
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.684139733953244e-05,
+      "loss": 15.8443,
+      "step": 12256
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.683760641912147e-05,
+      "loss": 17.9011,
+      "step": 12257
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.683381528198967e-05,
+      "loss": 14.3401,
+      "step": 12258
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.683002392816763e-05,
+      "loss": 15.4587,
+      "step": 12259
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.682623235768597e-05,
+      "loss": 15.1242,
+      "step": 12260
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.682244057057533e-05,
+      "loss": 14.9911,
+      "step": 12261
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.681864856686633e-05,
+      "loss": 14.8945,
+      "step": 12262
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.681485634658956e-05,
+      "loss": 15.6978,
+      "step": 12263
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.681106390977569e-05,
+      "loss": 14.5378,
+      "step": 12264
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.680727125645528e-05,
+      "loss": 14.8281,
+      "step": 12265
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.680347838665902e-05,
+      "loss": 16.0117,
+      "step": 12266
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.67996853004175e-05,
+      "loss": 16.8321,
+      "step": 12267
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.679589199776137e-05,
+      "loss": 17.3993,
+      "step": 12268
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.679209847872124e-05,
+      "loss": 14.9159,
+      "step": 12269
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.678830474332776e-05,
+      "loss": 15.3023,
+      "step": 12270
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.678451079161158e-05,
+      "loss": 14.9549,
+      "step": 12271
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.67807166236033e-05,
+      "loss": 15.6671,
+      "step": 12272
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.67769222393336e-05,
+      "loss": 14.9178,
+      "step": 12273
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.677312763883312e-05,
+      "loss": 17.5608,
+      "step": 12274
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.676933282213244e-05,
+      "loss": 15.4213,
+      "step": 12275
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.676553778926225e-05,
+      "loss": 14.9193,
+      "step": 12276
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.676174254025321e-05,
+      "loss": 17.0013,
+      "step": 12277
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.675794707513595e-05,
+      "loss": 15.9305,
+      "step": 12278
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.675415139394112e-05,
+      "loss": 14.9034,
+      "step": 12279
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.675035549669936e-05,
+      "loss": 14.9551,
+      "step": 12280
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.674655938344133e-05,
+      "loss": 17.5397,
+      "step": 12281
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.67427630541977e-05,
+      "loss": 15.3238,
+      "step": 12282
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.673896650899912e-05,
+      "loss": 16.5921,
+      "step": 12283
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.673516974787623e-05,
+      "loss": 15.3328,
+      "step": 12284
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.67313727708597e-05,
+      "loss": 15.0077,
+      "step": 12285
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.672757557798021e-05,
+      "loss": 14.0442,
+      "step": 12286
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.67237781692684e-05,
+      "loss": 16.0688,
+      "step": 12287
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.671998054475495e-05,
+      "loss": 17.1063,
+      "step": 12288
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.671618270447052e-05,
+      "loss": 14.3585,
+      "step": 12289
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.671238464844576e-05,
+      "loss": 16.6172,
+      "step": 12290
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.670858637671137e-05,
+      "loss": 15.5209,
+      "step": 12291
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.670478788929802e-05,
+      "loss": 13.5966,
+      "step": 12292
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.670098918623636e-05,
+      "loss": 14.9657,
+      "step": 12293
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.669719026755708e-05,
+      "loss": 15.5298,
+      "step": 12294
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.669339113329087e-05,
+      "loss": 14.9889,
+      "step": 12295
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.668959178346839e-05,
+      "loss": 15.9902,
+      "step": 12296
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.668579221812033e-05,
+      "loss": 15.8905,
+      "step": 12297
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.668199243727736e-05,
+      "loss": 16.3156,
+      "step": 12298
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.667819244097017e-05,
+      "loss": 16.3971,
+      "step": 12299
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.667439222922946e-05,
+      "loss": 16.6858,
+      "step": 12300
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.667059180208592e-05,
+      "loss": 15.9842,
+      "step": 12301
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.66667911595702e-05,
+      "loss": 15.9483,
+      "step": 12302
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.666299030171302e-05,
+      "loss": 14.9502,
+      "step": 12303
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.665918922854509e-05,
+      "loss": 15.1235,
+      "step": 12304
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.665538794009708e-05,
+      "loss": 17.1706,
+      "step": 12305
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.66515864363997e-05,
+      "loss": 16.3022,
+      "step": 12306
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.664778471748362e-05,
+      "loss": 15.3701,
+      "step": 12307
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.664398278337957e-05,
+      "loss": 16.416,
+      "step": 12308
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.664018063411824e-05,
+      "loss": 16.7682,
+      "step": 12309
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.663637826973035e-05,
+      "loss": 17.3934,
+      "step": 12310
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.663257569024658e-05,
+      "loss": 13.9996,
+      "step": 12311
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.662877289569764e-05,
+      "loss": 13.7571,
+      "step": 12312
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 7.662496988611426e-05,
+      "loss": 15.1765,
+      "step": 12313
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.662116666152712e-05,
+      "loss": 14.0806,
+      "step": 12314
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.661736322196695e-05,
+      "loss": 16.8184,
+      "step": 12315
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.661355956746447e-05,
+      "loss": 14.9674,
+      "step": 12316
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.660975569805037e-05,
+      "loss": 17.6943,
+      "step": 12317
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.660595161375541e-05,
+      "loss": 16.7666,
+      "step": 12318
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.660214731461028e-05,
+      "loss": 15.6157,
+      "step": 12319
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.659834280064571e-05,
+      "loss": 14.3444,
+      "step": 12320
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.659453807189239e-05,
+      "loss": 14.0905,
+      "step": 12321
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.659073312838109e-05,
+      "loss": 15.3508,
+      "step": 12322
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.65869279701425e-05,
+      "loss": 14.9882,
+      "step": 12323
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.658312259720739e-05,
+      "loss": 15.5758,
+      "step": 12324
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.657931700960646e-05,
+      "loss": 15.124,
+      "step": 12325
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.657551120737043e-05,
+      "loss": 14.174,
+      "step": 12326
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.657170519053006e-05,
+      "loss": 17.0872,
+      "step": 12327
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.656789895911605e-05,
+      "loss": 16.5092,
+      "step": 12328
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.65640925131592e-05,
+      "loss": 16.4005,
+      "step": 12329
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.656028585269018e-05,
+      "loss": 14.2836,
+      "step": 12330
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.655647897773977e-05,
+      "loss": 16.3842,
+      "step": 12331
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.655267188833868e-05,
+      "loss": 16.1801,
+      "step": 12332
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.65488645845177e-05,
+      "loss": 14.3646,
+      "step": 12333
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.65450570663075e-05,
+      "loss": 15.3624,
+      "step": 12334
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.654124933373892e-05,
+      "loss": 14.8994,
+      "step": 12335
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.653744138684263e-05,
+      "loss": 15.1335,
+      "step": 12336
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.653363322564944e-05,
+      "loss": 15.4629,
+      "step": 12337
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.652982485019005e-05,
+      "loss": 17.006,
+      "step": 12338
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.652601626049525e-05,
+      "loss": 15.5744,
+      "step": 12339
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.652220745659578e-05,
+      "loss": 16.026,
+      "step": 12340
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.65183984385224e-05,
+      "loss": 15.254,
+      "step": 12341
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.651458920630586e-05,
+      "loss": 15.7006,
+      "step": 12342
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.651077975997696e-05,
+      "loss": 16.8939,
+      "step": 12343
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.65069700995664e-05,
+      "loss": 16.4457,
+      "step": 12344
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.650316022510498e-05,
+      "loss": 15.5387,
+      "step": 12345
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.649935013662347e-05,
+      "loss": 16.3284,
+      "step": 12346
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.649553983415263e-05,
+      "loss": 17.8053,
+      "step": 12347
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.649172931772322e-05,
+      "loss": 15.5573,
+      "step": 12348
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.648791858736603e-05,
+      "loss": 15.7984,
+      "step": 12349
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.648410764311183e-05,
+      "loss": 15.7258,
+      "step": 12350
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.648029648499136e-05,
+      "loss": 16.0772,
+      "step": 12351
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.647648511303544e-05,
+      "loss": 15.6019,
+      "step": 12352
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.647267352727483e-05,
+      "loss": 16.748,
+      "step": 12353
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.64688617277403e-05,
+      "loss": 16.9475,
+      "step": 12354
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.646504971446264e-05,
+      "loss": 15.8818,
+      "step": 12355
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.646123748747265e-05,
+      "loss": 17.3824,
+      "step": 12356
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.645742504680109e-05,
+      "loss": 14.68,
+      "step": 12357
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.645361239247876e-05,
+      "loss": 14.5345,
+      "step": 12358
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.644979952453643e-05,
+      "loss": 17.1921,
+      "step": 12359
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.644598644300492e-05,
+      "loss": 16.0735,
+      "step": 12360
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.6442173147915e-05,
+      "loss": 16.0042,
+      "step": 12361
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.643835963929747e-05,
+      "loss": 15.9234,
+      "step": 12362
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.643454591718312e-05,
+      "loss": 14.92,
+      "step": 12363
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.643073198160276e-05,
+      "loss": 15.8785,
+      "step": 12364
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.642691783258716e-05,
+      "loss": 15.1876,
+      "step": 12365
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.642310347016715e-05,
+      "loss": 14.5642,
+      "step": 12366
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.641928889437353e-05,
+      "loss": 15.8108,
+      "step": 12367
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.641547410523709e-05,
+      "loss": 17.9016,
+      "step": 12368
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.641165910278862e-05,
+      "loss": 17.4425,
+      "step": 12369
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.640784388705898e-05,
+      "loss": 16.2931,
+      "step": 12370
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.640402845807892e-05,
+      "loss": 14.8056,
+      "step": 12371
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.64002128158793e-05,
+      "loss": 15.3865,
+      "step": 12372
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.63963969604909e-05,
+      "loss": 16.1398,
+      "step": 12373
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.639258089194454e-05,
+      "loss": 16.7527,
+      "step": 12374
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.638876461027104e-05,
+      "loss": 15.0114,
+      "step": 12375
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.638494811550121e-05,
+      "loss": 16.6466,
+      "step": 12376
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.638113140766589e-05,
+      "loss": 15.9611,
+      "step": 12377
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.637731448679588e-05,
+      "loss": 16.7416,
+      "step": 12378
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.637349735292201e-05,
+      "loss": 14.9839,
+      "step": 12379
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.636968000607509e-05,
+      "loss": 16.2718,
+      "step": 12380
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.636586244628598e-05,
+      "loss": 14.5031,
+      "step": 12381
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.636204467358548e-05,
+      "loss": 16.8859,
+      "step": 12382
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.635822668800444e-05,
+      "loss": 14.7925,
+      "step": 12383
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.635440848957366e-05,
+      "loss": 15.5956,
+      "step": 12384
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.635059007832401e-05,
+      "loss": 15.3956,
+      "step": 12385
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.63467714542863e-05,
+      "loss": 14.8877,
+      "step": 12386
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.634295261749135e-05,
+      "loss": 17.0352,
+      "step": 12387
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.633913356797004e-05,
+      "loss": 15.5236,
+      "step": 12388
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.63353143057532e-05,
+      "loss": 16.0758,
+      "step": 12389
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.633149483087166e-05,
+      "loss": 15.7308,
+      "step": 12390
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.632767514335625e-05,
+      "loss": 16.1993,
+      "step": 12391
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.632385524323784e-05,
+      "loss": 18.0668,
+      "step": 12392
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.632003513054726e-05,
+      "loss": 18.1021,
+      "step": 12393
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.631621480531537e-05,
+      "loss": 15.7154,
+      "step": 12394
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.631239426757301e-05,
+      "loss": 15.5277,
+      "step": 12395
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.630857351735106e-05,
+      "loss": 16.8654,
+      "step": 12396
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.630475255468034e-05,
+      "loss": 14.9507,
+      "step": 12397
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.630093137959171e-05,
+      "loss": 16.4466,
+      "step": 12398
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.629710999211603e-05,
+      "loss": 16.4858,
+      "step": 12399
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.629328839228418e-05,
+      "loss": 16.6788,
+      "step": 12400
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.628946658012699e-05,
+      "loss": 16.1478,
+      "step": 12401
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.628564455567535e-05,
+      "loss": 15.1591,
+      "step": 12402
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.628182231896009e-05,
+      "loss": 16.4098,
+      "step": 12403
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.627799987001209e-05,
+      "loss": 15.3901,
+      "step": 12404
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.627417720886223e-05,
+      "loss": 16.938,
+      "step": 12405
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.627035433554138e-05,
+      "loss": 16.3696,
+      "step": 12406
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.626653125008039e-05,
+      "loss": 16.6836,
+      "step": 12407
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.626270795251016e-05,
+      "loss": 15.5866,
+      "step": 12408
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.625888444286152e-05,
+      "loss": 16.4316,
+      "step": 12409
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.62550607211654e-05,
+      "loss": 15.4939,
+      "step": 12410
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.625123678745263e-05,
+      "loss": 16.2886,
+      "step": 12411
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.624741264175413e-05,
+      "loss": 16.5157,
+      "step": 12412
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.624358828410075e-05,
+      "loss": 15.4053,
+      "step": 12413
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.62397637145234e-05,
+      "loss": 15.9404,
+      "step": 12414
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.623593893305295e-05,
+      "loss": 15.4091,
+      "step": 12415
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.623211393972025e-05,
+      "loss": 16.6027,
+      "step": 12416
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.622828873455625e-05,
+      "loss": 17.3478,
+      "step": 12417
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.62244633175918e-05,
+      "loss": 16.0568,
+      "step": 12418
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.622063768885783e-05,
+      "loss": 14.6549,
+      "step": 12419
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.621681184838517e-05,
+      "loss": 13.9873,
+      "step": 12420
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.621298579620477e-05,
+      "loss": 14.8896,
+      "step": 12421
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.620915953234751e-05,
+      "loss": 16.5665,
+      "step": 12422
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.620533305684429e-05,
+      "loss": 16.0721,
+      "step": 12423
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.620150636972599e-05,
+      "loss": 17.1576,
+      "step": 12424
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.619767947102354e-05,
+      "loss": 16.0935,
+      "step": 12425
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.619385236076782e-05,
+      "loss": 15.7863,
+      "step": 12426
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.619002503898975e-05,
+      "loss": 14.2837,
+      "step": 12427
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.618619750572023e-05,
+      "loss": 16.8067,
+      "step": 12428
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.618236976099017e-05,
+      "loss": 15.0068,
+      "step": 12429
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.617854180483048e-05,
+      "loss": 17.2673,
+      "step": 12430
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.617471363727209e-05,
+      "loss": 15.0289,
+      "step": 12431
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.617088525834587e-05,
+      "loss": 16.1019,
+      "step": 12432
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 7.616705666808277e-05,
+      "loss": 15.8974,
+      "step": 12433
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.616322786651371e-05,
+      "loss": 14.8196,
+      "step": 12434
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.615939885366957e-05,
+      "loss": 16.371,
+      "step": 12435
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.615556962958133e-05,
+      "loss": 15.5398,
+      "step": 12436
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.615174019427983e-05,
+      "loss": 14.7725,
+      "step": 12437
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.614791054779609e-05,
+      "loss": 16.4688,
+      "step": 12438
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.614408069016097e-05,
+      "loss": 14.2002,
+      "step": 12439
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.614025062140542e-05,
+      "loss": 17.389,
+      "step": 12440
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.613642034156035e-05,
+      "loss": 15.1199,
+      "step": 12441
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.613258985065672e-05,
+      "loss": 15.4686,
+      "step": 12442
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.612875914872544e-05,
+      "loss": 15.2625,
+      "step": 12443
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.612492823579745e-05,
+      "loss": 14.0747,
+      "step": 12444
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.612109711190367e-05,
+      "loss": 17.8927,
+      "step": 12445
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.611726577707509e-05,
+      "loss": 16.5629,
+      "step": 12446
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.611343423134258e-05,
+      "loss": 15.8252,
+      "step": 12447
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.610960247473713e-05,
+      "loss": 16.604,
+      "step": 12448
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.610577050728965e-05,
+      "loss": 15.9248,
+      "step": 12449
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.610193832903112e-05,
+      "loss": 16.7284,
+      "step": 12450
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.609810593999244e-05,
+      "loss": 16.2963,
+      "step": 12451
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.609427334020462e-05,
+      "loss": 16.8305,
+      "step": 12452
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.609044052969853e-05,
+      "loss": 15.276,
+      "step": 12453
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.60866075085052e-05,
+      "loss": 16.8678,
+      "step": 12454
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.608277427665553e-05,
+      "loss": 14.5575,
+      "step": 12455
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.60789408341805e-05,
+      "loss": 14.0317,
+      "step": 12456
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.607510718111105e-05,
+      "loss": 18.3025,
+      "step": 12457
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.607127331747814e-05,
+      "loss": 15.8969,
+      "step": 12458
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.606743924331274e-05,
+      "loss": 14.9021,
+      "step": 12459
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.606360495864579e-05,
+      "loss": 13.8283,
+      "step": 12460
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.605977046350828e-05,
+      "loss": 17.851,
+      "step": 12461
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.605593575793116e-05,
+      "loss": 16.0044,
+      "step": 12462
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.60521008419454e-05,
+      "loss": 14.3343,
+      "step": 12463
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.604826571558197e-05,
+      "loss": 14.8188,
+      "step": 12464
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.604443037887183e-05,
+      "loss": 17.5707,
+      "step": 12465
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.604059483184595e-05,
+      "loss": 16.4182,
+      "step": 12466
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.603675907453534e-05,
+      "loss": 14.935,
+      "step": 12467
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.603292310697091e-05,
+      "loss": 16.8424,
+      "step": 12468
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.602908692918371e-05,
+      "loss": 16.5365,
+      "step": 12469
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.602525054120465e-05,
+      "loss": 15.5464,
+      "step": 12470
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.602141394306476e-05,
+      "loss": 16.435,
+      "step": 12471
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.601757713479498e-05,
+      "loss": 14.0032,
+      "step": 12472
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.601374011642634e-05,
+      "loss": 15.8715,
+      "step": 12473
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.600990288798977e-05,
+      "loss": 15.3113,
+      "step": 12474
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.600606544951632e-05,
+      "loss": 16.7813,
+      "step": 12475
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.600222780103691e-05,
+      "loss": 16.7205,
+      "step": 12476
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.599838994258259e-05,
+      "loss": 17.8161,
+      "step": 12477
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.599455187418432e-05,
+      "loss": 17.7867,
+      "step": 12478
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.59907135958731e-05,
+      "loss": 14.7351,
+      "step": 12479
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.598687510767991e-05,
+      "loss": 15.9581,
+      "step": 12480
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.598303640963577e-05,
+      "loss": 13.6959,
+      "step": 12481
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.597919750177168e-05,
+      "loss": 15.3305,
+      "step": 12482
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.597535838411862e-05,
+      "loss": 14.7368,
+      "step": 12483
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.597151905670762e-05,
+      "loss": 17.2532,
+      "step": 12484
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.596767951956964e-05,
+      "loss": 16.619,
+      "step": 12485
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.596383977273573e-05,
+      "loss": 14.3492,
+      "step": 12486
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.595999981623688e-05,
+      "loss": 17.7673,
+      "step": 12487
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.595615965010409e-05,
+      "loss": 17.3848,
+      "step": 12488
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.595231927436837e-05,
+      "loss": 15.0018,
+      "step": 12489
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.594847868906076e-05,
+      "loss": 15.7767,
+      "step": 12490
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.594463789421224e-05,
+      "loss": 14.9835,
+      "step": 12491
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.594079688985385e-05,
+      "loss": 17.0832,
+      "step": 12492
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.593695567601658e-05,
+      "loss": 15.3878,
+      "step": 12493
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.593311425273147e-05,
+      "loss": 17.2377,
+      "step": 12494
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.592927262002954e-05,
+      "loss": 15.8994,
+      "step": 12495
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.59254307779418e-05,
+      "loss": 15.1182,
+      "step": 12496
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.592158872649928e-05,
+      "loss": 15.9506,
+      "step": 12497
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.591774646573301e-05,
+      "loss": 16.4041,
+      "step": 12498
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.591390399567402e-05,
+      "loss": 15.3744,
+      "step": 12499
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.591006131635333e-05,
+      "loss": 16.2027,
+      "step": 12500
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.590621842780198e-05,
+      "loss": 15.5133,
+      "step": 12501
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.590237533005098e-05,
+      "loss": 15.3586,
+      "step": 12502
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.58985320231314e-05,
+      "loss": 16.1036,
+      "step": 12503
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.589468850707425e-05,
+      "loss": 17.373,
+      "step": 12504
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.589084478191056e-05,
+      "loss": 15.6139,
+      "step": 12505
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.588700084767138e-05,
+      "loss": 16.9186,
+      "step": 12506
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.588315670438777e-05,
+      "loss": 15.5094,
+      "step": 12507
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.587931235209074e-05,
+      "loss": 15.4775,
+      "step": 12508
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.587546779081136e-05,
+      "loss": 17.0455,
+      "step": 12509
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.587162302058066e-05,
+      "loss": 13.9493,
+      "step": 12510
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.58677780414297e-05,
+      "loss": 16.1734,
+      "step": 12511
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.586393285338953e-05,
+      "loss": 14.3743,
+      "step": 12512
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.586008745649118e-05,
+      "loss": 16.5329,
+      "step": 12513
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.585624185076572e-05,
+      "loss": 15.2099,
+      "step": 12514
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.585239603624421e-05,
+      "loss": 16.2405,
+      "step": 12515
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.584855001295767e-05,
+      "loss": 15.6665,
+      "step": 12516
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.584470378093722e-05,
+      "loss": 14.5275,
+      "step": 12517
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.584085734021384e-05,
+      "loss": 17.0153,
+      "step": 12518
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.583701069081867e-05,
+      "loss": 15.9629,
+      "step": 12519
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.583316383278273e-05,
+      "loss": 15.2326,
+      "step": 12520
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.582931676613709e-05,
+      "loss": 17.262,
+      "step": 12521
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.582546949091281e-05,
+      "loss": 15.355,
+      "step": 12522
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.582162200714098e-05,
+      "loss": 14.9671,
+      "step": 12523
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.581777431485264e-05,
+      "loss": 16.6295,
+      "step": 12524
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.581392641407887e-05,
+      "loss": 16.2751,
+      "step": 12525
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.581007830485076e-05,
+      "loss": 17.1329,
+      "step": 12526
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.580622998719937e-05,
+      "loss": 13.5799,
+      "step": 12527
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.58023814611558e-05,
+      "loss": 15.7619,
+      "step": 12528
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.579853272675108e-05,
+      "loss": 14.0072,
+      "step": 12529
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.579468378401633e-05,
+      "loss": 16.6165,
+      "step": 12530
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.57908346329826e-05,
+      "loss": 17.3736,
+      "step": 12531
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.578698527368102e-05,
+      "loss": 15.5647,
+      "step": 12532
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.578313570614262e-05,
+      "loss": 15.6755,
+      "step": 12533
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.577928593039853e-05,
+      "loss": 16.0415,
+      "step": 12534
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.577543594647981e-05,
+      "loss": 16.5871,
+      "step": 12535
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.577158575441756e-05,
+      "loss": 14.2333,
+      "step": 12536
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.576773535424288e-05,
+      "loss": 16.9064,
+      "step": 12537
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.576388474598685e-05,
+      "loss": 14.7566,
+      "step": 12538
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.576003392968056e-05,
+      "loss": 15.9814,
+      "step": 12539
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.575618290535514e-05,
+      "loss": 15.2439,
+      "step": 12540
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.575233167304163e-05,
+      "loss": 14.4744,
+      "step": 12541
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.574848023277117e-05,
+      "loss": 16.3937,
+      "step": 12542
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.574462858457488e-05,
+      "loss": 18.8139,
+      "step": 12543
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.574077672848381e-05,
+      "loss": 17.3879,
+      "step": 12544
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.57369246645291e-05,
+      "loss": 14.9923,
+      "step": 12545
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.573307239274185e-05,
+      "loss": 15.6928,
+      "step": 12546
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.572921991315316e-05,
+      "loss": 16.2234,
+      "step": 12547
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.572536722579415e-05,
+      "loss": 13.3448,
+      "step": 12548
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.572151433069594e-05,
+      "loss": 16.0894,
+      "step": 12549
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.571766122788962e-05,
+      "loss": 15.2808,
+      "step": 12550
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.571380791740633e-05,
+      "loss": 15.9444,
+      "step": 12551
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.570995439927715e-05,
+      "loss": 16.5336,
+      "step": 12552
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.570610067353323e-05,
+      "loss": 17.4481,
+      "step": 12553
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.570224674020567e-05,
+      "loss": 14.5132,
+      "step": 12554
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.569839259932563e-05,
+      "loss": 17.4751,
+      "step": 12555
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.569453825092419e-05,
+      "loss": 17.7302,
+      "step": 12556
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.569068369503249e-05,
+      "loss": 15.594,
+      "step": 12557
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.568682893168164e-05,
+      "loss": 14.814,
+      "step": 12558
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.56829739609028e-05,
+      "loss": 15.3225,
+      "step": 12559
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.567911878272708e-05,
+      "loss": 14.8626,
+      "step": 12560
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.567526339718562e-05,
+      "loss": 14.8686,
+      "step": 12561
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.567140780430956e-05,
+      "loss": 16.2427,
+      "step": 12562
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.566755200413002e-05,
+      "loss": 15.0938,
+      "step": 12563
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.566369599667813e-05,
+      "loss": 15.8413,
+      "step": 12564
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.565983978198506e-05,
+      "loss": 16.0406,
+      "step": 12565
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.56559833600819e-05,
+      "loss": 15.5705,
+      "step": 12566
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.565212673099984e-05,
+      "loss": 15.1542,
+      "step": 12567
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.564826989476999e-05,
+      "loss": 16.4479,
+      "step": 12568
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.564441285142354e-05,
+      "loss": 15.875,
+      "step": 12569
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.564055560099157e-05,
+      "loss": 15.9609,
+      "step": 12570
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.563669814350529e-05,
+      "loss": 16.0434,
+      "step": 12571
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.563284047899582e-05,
+      "loss": 16.6278,
+      "step": 12572
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.562898260749429e-05,
+      "loss": 17.2294,
+      "step": 12573
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.562512452903192e-05,
+      "loss": 15.5349,
+      "step": 12574
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.562126624363979e-05,
+      "loss": 15.3944,
+      "step": 12575
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.561740775134912e-05,
+      "loss": 16.794,
+      "step": 12576
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.561354905219102e-05,
+      "loss": 17.1505,
+      "step": 12577
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.560969014619668e-05,
+      "loss": 17.3157,
+      "step": 12578
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.560583103339724e-05,
+      "loss": 15.4281,
+      "step": 12579
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.560197171382388e-05,
+      "loss": 16.4123,
+      "step": 12580
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.559811218750777e-05,
+      "loss": 15.5305,
+      "step": 12581
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.559425245448006e-05,
+      "loss": 16.0264,
+      "step": 12582
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.559039251477192e-05,
+      "loss": 15.4188,
+      "step": 12583
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.558653236841452e-05,
+      "loss": 15.4243,
+      "step": 12584
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.558267201543905e-05,
+      "loss": 15.371,
+      "step": 12585
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.557881145587667e-05,
+      "loss": 16.4273,
+      "step": 12586
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.557495068975854e-05,
+      "loss": 14.1177,
+      "step": 12587
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.557108971711586e-05,
+      "loss": 14.7312,
+      "step": 12588
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.556722853797979e-05,
+      "loss": 16.1531,
+      "step": 12589
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.556336715238153e-05,
+      "loss": 15.4358,
+      "step": 12590
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.555950556035226e-05,
+      "loss": 14.8097,
+      "step": 12591
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.555564376192314e-05,
+      "loss": 14.9908,
+      "step": 12592
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.555178175712538e-05,
+      "loss": 14.5646,
+      "step": 12593
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.554791954599015e-05,
+      "loss": 15.7083,
+      "step": 12594
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.554405712854866e-05,
+      "loss": 17.3226,
+      "step": 12595
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.554019450483208e-05,
+      "loss": 15.2375,
+      "step": 12596
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.55363316748716e-05,
+      "loss": 14.2488,
+      "step": 12597
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.553246863869841e-05,
+      "loss": 17.0881,
+      "step": 12598
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.552860539634373e-05,
+      "loss": 15.279,
+      "step": 12599
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.552474194783874e-05,
+      "loss": 16.7914,
+      "step": 12600
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.552087829321464e-05,
+      "loss": 16.4287,
+      "step": 12601
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.551701443250263e-05,
+      "loss": 16.7286,
+      "step": 12602
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.551315036573392e-05,
+      "loss": 14.8917,
+      "step": 12603
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.550928609293967e-05,
+      "loss": 13.7204,
+      "step": 12604
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.550542161415117e-05,
+      "loss": 15.7613,
+      "step": 12605
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.550155692939955e-05,
+      "loss": 16.8125,
+      "step": 12606
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.549769203871606e-05,
+      "loss": 14.6795,
+      "step": 12607
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.549382694213188e-05,
+      "loss": 15.6322,
+      "step": 12608
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.548996163967826e-05,
+      "loss": 15.8045,
+      "step": 12609
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.548609613138637e-05,
+      "loss": 17.3939,
+      "step": 12610
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.548223041728746e-05,
+      "loss": 13.8339,
+      "step": 12611
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.547836449741272e-05,
+      "loss": 15.7906,
+      "step": 12612
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.54744983717934e-05,
+      "loss": 15.9359,
+      "step": 12613
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.547063204046068e-05,
+      "loss": 16.4219,
+      "step": 12614
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.54667655034458e-05,
+      "loss": 16.4275,
+      "step": 12615
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.546289876078e-05,
+      "loss": 14.4351,
+      "step": 12616
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.545903181249449e-05,
+      "loss": 15.1938,
+      "step": 12617
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.545516465862048e-05,
+      "loss": 18.7095,
+      "step": 12618
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.545129729918923e-05,
+      "loss": 16.3321,
+      "step": 12619
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.544742973423196e-05,
+      "loss": 15.5883,
+      "step": 12620
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.544356196377989e-05,
+      "loss": 16.0786,
+      "step": 12621
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.543969398786427e-05,
+      "loss": 15.5912,
+      "step": 12622
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.543582580651631e-05,
+      "loss": 15.0774,
+      "step": 12623
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.543195741976728e-05,
+      "loss": 14.8164,
+      "step": 12624
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.542808882764839e-05,
+      "loss": 15.6284,
+      "step": 12625
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.54242200301909e-05,
+      "loss": 15.3084,
+      "step": 12626
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.542035102742602e-05,
+      "loss": 15.5594,
+      "step": 12627
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.541648181938503e-05,
+      "loss": 17.2183,
+      "step": 12628
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.541261240609917e-05,
+      "loss": 16.5789,
+      "step": 12629
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.540874278759968e-05,
+      "loss": 15.1106,
+      "step": 12630
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.540487296391779e-05,
+      "loss": 14.3725,
+      "step": 12631
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.540100293508476e-05,
+      "loss": 16.7032,
+      "step": 12632
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.539713270113185e-05,
+      "loss": 16.699,
+      "step": 12633
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.539326226209031e-05,
+      "loss": 16.5171,
+      "step": 12634
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.538939161799139e-05,
+      "loss": 16.3393,
+      "step": 12635
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.538552076886635e-05,
+      "loss": 16.1366,
+      "step": 12636
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.538164971474646e-05,
+      "loss": 17.686,
+      "step": 12637
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.537777845566296e-05,
+      "loss": 15.8512,
+      "step": 12638
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.537390699164712e-05,
+      "loss": 16.8003,
+      "step": 12639
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.537003532273022e-05,
+      "loss": 15.2526,
+      "step": 12640
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.536616344894349e-05,
+      "loss": 19.0782,
+      "step": 12641
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.536229137031821e-05,
+      "loss": 15.0851,
+      "step": 12642
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.535841908688568e-05,
+      "loss": 16.5183,
+      "step": 12643
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.535454659867711e-05,
+      "loss": 18.3235,
+      "step": 12644
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.535067390572383e-05,
+      "loss": 15.3066,
+      "step": 12645
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.534680100805706e-05,
+      "loss": 15.3175,
+      "step": 12646
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.534292790570812e-05,
+      "loss": 17.5478,
+      "step": 12647
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.533905459870826e-05,
+      "loss": 16.7938,
+      "step": 12648
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.533518108708877e-05,
+      "loss": 16.5525,
+      "step": 12649
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.533130737088092e-05,
+      "loss": 15.4606,
+      "step": 12650
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.5327433450116e-05,
+      "loss": 17.0062,
+      "step": 12651
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.532355932482529e-05,
+      "loss": 15.4836,
+      "step": 12652
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.531968499504008e-05,
+      "loss": 17.2302,
+      "step": 12653
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.531581046079164e-05,
+      "loss": 14.5116,
+      "step": 12654
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.531193572211127e-05,
+      "loss": 14.8619,
+      "step": 12655
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.530806077903025e-05,
+      "loss": 16.8252,
+      "step": 12656
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.53041856315799e-05,
+      "loss": 15.9076,
+      "step": 12657
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.530031027979147e-05,
+      "loss": 16.5484,
+      "step": 12658
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.52964347236963e-05,
+      "loss": 13.7576,
+      "step": 12659
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.529255896332563e-05,
+      "loss": 14.7732,
+      "step": 12660
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.528868299871081e-05,
+      "loss": 14.9683,
+      "step": 12661
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.528480682988313e-05,
+      "loss": 15.885,
+      "step": 12662
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.528093045687387e-05,
+      "loss": 16.5114,
+      "step": 12663
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.527705387971433e-05,
+      "loss": 16.0567,
+      "step": 12664
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.527317709843585e-05,
+      "loss": 15.5735,
+      "step": 12665
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.526930011306969e-05,
+      "loss": 16.1283,
+      "step": 12666
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.52654229236472e-05,
+      "loss": 15.1448,
+      "step": 12667
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.526154553019968e-05,
+      "loss": 14.4836,
+      "step": 12668
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.52576679327584e-05,
+      "loss": 16.0283,
+      "step": 12669
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.525379013135475e-05,
+      "loss": 14.9346,
+      "step": 12670
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.524991212601995e-05,
+      "loss": 16.8941,
+      "step": 12671
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.524603391678541e-05,
+      "loss": 15.0605,
+      "step": 12672
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 7.524215550368237e-05,
+      "loss": 14.3466,
+      "step": 12673
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.52382768867422e-05,
+      "loss": 15.8257,
+      "step": 12674
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.523439806599619e-05,
+      "loss": 16.7136,
+      "step": 12675
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.523051904147569e-05,
+      "loss": 15.9205,
+      "step": 12676
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.522663981321199e-05,
+      "loss": 15.0759,
+      "step": 12677
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.522276038123643e-05,
+      "loss": 16.156,
+      "step": 12678
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.521888074558037e-05,
+      "loss": 14.9672,
+      "step": 12679
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.521500090627509e-05,
+      "loss": 16.6281,
+      "step": 12680
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.521112086335194e-05,
+      "loss": 15.3177,
+      "step": 12681
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.520724061684227e-05,
+      "loss": 15.6869,
+      "step": 12682
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.52033601667774e-05,
+      "loss": 14.7973,
+      "step": 12683
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.519947951318865e-05,
+      "loss": 15.791,
+      "step": 12684
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.519559865610738e-05,
+      "loss": 13.5381,
+      "step": 12685
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.519171759556491e-05,
+      "loss": 13.6472,
+      "step": 12686
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.518783633159261e-05,
+      "loss": 15.7178,
+      "step": 12687
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.518395486422178e-05,
+      "loss": 16.7094,
+      "step": 12688
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.518007319348381e-05,
+      "loss": 18.3406,
+      "step": 12689
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.517619131941002e-05,
+      "loss": 14.6118,
+      "step": 12690
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.517230924203175e-05,
+      "loss": 17.6626,
+      "step": 12691
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.516842696138035e-05,
+      "loss": 14.1994,
+      "step": 12692
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.516454447748721e-05,
+      "loss": 15.9178,
+      "step": 12693
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.516066179038363e-05,
+      "loss": 16.3527,
+      "step": 12694
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.5156778900101e-05,
+      "loss": 14.7258,
+      "step": 12695
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.515289580667064e-05,
+      "loss": 16.1168,
+      "step": 12696
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.514901251012394e-05,
+      "loss": 16.2452,
+      "step": 12697
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.514512901049224e-05,
+      "loss": 14.8456,
+      "step": 12698
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.514124530780691e-05,
+      "loss": 15.332,
+      "step": 12699
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.513736140209931e-05,
+      "loss": 15.4684,
+      "step": 12700
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.51334772934008e-05,
+      "loss": 14.8258,
+      "step": 12701
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.512959298174273e-05,
+      "loss": 14.8934,
+      "step": 12702
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.512570846715652e-05,
+      "loss": 14.446,
+      "step": 12703
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.512182374967347e-05,
+      "loss": 16.2197,
+      "step": 12704
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.511793882932501e-05,
+      "loss": 13.7851,
+      "step": 12705
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.511405370614247e-05,
+      "loss": 15.8641,
+      "step": 12706
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.511016838015723e-05,
+      "loss": 17.1573,
+      "step": 12707
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.510628285140069e-05,
+      "loss": 16.7391,
+      "step": 12708
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.51023971199042e-05,
+      "loss": 14.0164,
+      "step": 12709
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.509851118569915e-05,
+      "loss": 14.7946,
+      "step": 12710
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.50946250488169e-05,
+      "loss": 14.999,
+      "step": 12711
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.509073870928888e-05,
+      "loss": 16.3663,
+      "step": 12712
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.508685216714641e-05,
+      "loss": 15.5359,
+      "step": 12713
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.508296542242095e-05,
+      "loss": 16.5646,
+      "step": 12714
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.507907847514381e-05,
+      "loss": 14.736,
+      "step": 12715
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.507519132534644e-05,
+      "loss": 16.7013,
+      "step": 12716
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.507130397306017e-05,
+      "loss": 15.9421,
+      "step": 12717
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.506741641831645e-05,
+      "loss": 15.44,
+      "step": 12718
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.506352866114663e-05,
+      "loss": 16.2658,
+      "step": 12719
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.505964070158214e-05,
+      "loss": 16.6782,
+      "step": 12720
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.505575253965434e-05,
+      "loss": 14.7538,
+      "step": 12721
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.505186417539465e-05,
+      "loss": 16.3534,
+      "step": 12722
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.504797560883446e-05,
+      "loss": 15.7989,
+      "step": 12723
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.504408684000517e-05,
+      "loss": 15.9534,
+      "step": 12724
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.50401978689382e-05,
+      "loss": 15.5364,
+      "step": 12725
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.503630869566494e-05,
+      "loss": 16.657,
+      "step": 12726
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.503241932021681e-05,
+      "loss": 16.3318,
+      "step": 12727
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.502852974262519e-05,
+      "loss": 16.6117,
+      "step": 12728
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.502463996292151e-05,
+      "loss": 18.2977,
+      "step": 12729
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.502074998113718e-05,
+      "loss": 16.5676,
+      "step": 12730
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.501685979730361e-05,
+      "loss": 15.3143,
+      "step": 12731
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.501296941145223e-05,
+      "loss": 15.1162,
+      "step": 12732
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.500907882361444e-05,
+      "loss": 15.8305,
+      "step": 12733
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.500518803382164e-05,
+      "loss": 14.7508,
+      "step": 12734
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.500129704210526e-05,
+      "loss": 14.085,
+      "step": 12735
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.499740584849675e-05,
+      "loss": 16.5018,
+      "step": 12736
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.49935144530275e-05,
+      "loss": 16.6891,
+      "step": 12737
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.498962285572894e-05,
+      "loss": 15.9475,
+      "step": 12738
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.498573105663251e-05,
+      "loss": 14.933,
+      "step": 12739
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.498183905576963e-05,
+      "loss": 16.7818,
+      "step": 12740
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.497794685317173e-05,
+      "loss": 14.7703,
+      "step": 12741
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.497405444887022e-05,
+      "loss": 16.0596,
+      "step": 12742
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.497016184289656e-05,
+      "loss": 15.2778,
+      "step": 12743
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.496626903528217e-05,
+      "loss": 15.6601,
+      "step": 12744
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.49623760260585e-05,
+      "loss": 15.8335,
+      "step": 12745
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.495848281525694e-05,
+      "loss": 15.9339,
+      "step": 12746
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.495458940290901e-05,
+      "loss": 14.451,
+      "step": 12747
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.495069578904608e-05,
+      "loss": 15.812,
+      "step": 12748
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.494680197369962e-05,
+      "loss": 13.147,
+      "step": 12749
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.494290795690106e-05,
+      "loss": 16.1282,
+      "step": 12750
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.493901373868189e-05,
+      "loss": 14.874,
+      "step": 12751
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.493511931907349e-05,
+      "loss": 17.2101,
+      "step": 12752
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.493122469810735e-05,
+      "loss": 14.9904,
+      "step": 12753
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.49273298758149e-05,
+      "loss": 16.1591,
+      "step": 12754
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.492343485222762e-05,
+      "loss": 16.2365,
+      "step": 12755
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.491953962737693e-05,
+      "loss": 17.3276,
+      "step": 12756
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.491564420129432e-05,
+      "loss": 16.624,
+      "step": 12757
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.491174857401122e-05,
+      "loss": 16.2666,
+      "step": 12758
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.49078527455591e-05,
+      "loss": 14.3271,
+      "step": 12759
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.49039567159694e-05,
+      "loss": 15.7176,
+      "step": 12760
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.49000604852736e-05,
+      "loss": 15.9327,
+      "step": 12761
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.489616405350319e-05,
+      "loss": 15.0856,
+      "step": 12762
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.489226742068957e-05,
+      "loss": 16.0751,
+      "step": 12763
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.488837058686425e-05,
+      "loss": 13.9747,
+      "step": 12764
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.48844735520587e-05,
+      "loss": 14.1355,
+      "step": 12765
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.488057631630437e-05,
+      "loss": 15.2221,
+      "step": 12766
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.487667887963273e-05,
+      "loss": 15.0161,
+      "step": 12767
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.487278124207529e-05,
+      "loss": 16.3173,
+      "step": 12768
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.486888340366347e-05,
+      "loss": 17.02,
+      "step": 12769
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.48649853644288e-05,
+      "loss": 15.8691,
+      "step": 12770
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.48610871244027e-05,
+      "loss": 15.6406,
+      "step": 12771
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.48571886836167e-05,
+      "loss": 15.7609,
+      "step": 12772
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.485329004210227e-05,
+      "loss": 15.0872,
+      "step": 12773
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.484939119989086e-05,
+      "loss": 15.7423,
+      "step": 12774
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.4845492157014e-05,
+      "loss": 17.2726,
+      "step": 12775
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.484159291350315e-05,
+      "loss": 17.3275,
+      "step": 12776
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.48376934693898e-05,
+      "loss": 14.9698,
+      "step": 12777
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.483379382470543e-05,
+      "loss": 16.0396,
+      "step": 12778
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.482989397948155e-05,
+      "loss": 19.3344,
+      "step": 12779
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.482599393374962e-05,
+      "loss": 15.7789,
+      "step": 12780
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.48220936875412e-05,
+      "loss": 16.825,
+      "step": 12781
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.481819324088772e-05,
+      "loss": 15.6393,
+      "step": 12782
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.48142925938207e-05,
+      "loss": 15.9235,
+      "step": 12783
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.481039174637164e-05,
+      "loss": 15.0092,
+      "step": 12784
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.480649069857205e-05,
+      "loss": 15.1003,
+      "step": 12785
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.48025894504534e-05,
+      "loss": 16.2637,
+      "step": 12786
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.479868800204723e-05,
+      "loss": 16.7462,
+      "step": 12787
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.479478635338504e-05,
+      "loss": 15.6971,
+      "step": 12788
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.479088450449831e-05,
+      "loss": 15.7273,
+      "step": 12789
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.478698245541856e-05,
+      "loss": 16.9205,
+      "step": 12790
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.478308020617733e-05,
+      "loss": 14.4718,
+      "step": 12791
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.477917775680608e-05,
+      "loss": 16.4318,
+      "step": 12792
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.477527510733637e-05,
+      "loss": 16.6039,
+      "step": 12793
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.477137225779969e-05,
+      "loss": 17.8007,
+      "step": 12794
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.476746920822757e-05,
+      "loss": 16.158,
+      "step": 12795
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.47635659586515e-05,
+      "loss": 14.5938,
+      "step": 12796
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.475966250910303e-05,
+      "loss": 17.0925,
+      "step": 12797
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.475575885961368e-05,
+      "loss": 15.2903,
+      "step": 12798
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.475185501021496e-05,
+      "loss": 17.2704,
+      "step": 12799
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.47479509609384e-05,
+      "loss": 17.5879,
+      "step": 12800
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.474404671181552e-05,
+      "loss": 16.3575,
+      "step": 12801
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.474014226287785e-05,
+      "loss": 16.7302,
+      "step": 12802
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.473623761415693e-05,
+      "loss": 16.1911,
+      "step": 12803
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.473233276568429e-05,
+      "loss": 16.8655,
+      "step": 12804
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.472842771749144e-05,
+      "loss": 16.2033,
+      "step": 12805
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.472452246960993e-05,
+      "loss": 14.5231,
+      "step": 12806
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.472061702207131e-05,
+      "loss": 16.1297,
+      "step": 12807
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.47167113749071e-05,
+      "loss": 15.8169,
+      "step": 12808
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.471280552814885e-05,
+      "loss": 15.706,
+      "step": 12809
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.470889948182808e-05,
+      "loss": 17.0712,
+      "step": 12810
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.470499323597636e-05,
+      "loss": 14.3149,
+      "step": 12811
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.470108679062521e-05,
+      "loss": 16.6534,
+      "step": 12812
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.469718014580618e-05,
+      "loss": 14.9049,
+      "step": 12813
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.469327330155085e-05,
+      "loss": 16.1676,
+      "step": 12814
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.468936625789072e-05,
+      "loss": 15.406,
+      "step": 12815
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.468545901485737e-05,
+      "loss": 13.329,
+      "step": 12816
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.468155157248233e-05,
+      "loss": 15.0948,
+      "step": 12817
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.467764393079718e-05,
+      "loss": 16.7624,
+      "step": 12818
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.467373608983346e-05,
+      "loss": 15.1389,
+      "step": 12819
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.466982804962274e-05,
+      "loss": 14.0334,
+      "step": 12820
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.466591981019656e-05,
+      "loss": 15.3195,
+      "step": 12821
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.466201137158648e-05,
+      "loss": 15.2348,
+      "step": 12822
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.46581027338241e-05,
+      "loss": 15.4908,
+      "step": 12823
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.465419389694092e-05,
+      "loss": 14.7722,
+      "step": 12824
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.465028486096857e-05,
+      "loss": 16.8525,
+      "step": 12825
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.464637562593856e-05,
+      "loss": 14.5689,
+      "step": 12826
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.464246619188248e-05,
+      "loss": 15.8276,
+      "step": 12827
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.463855655883191e-05,
+      "loss": 14.8738,
+      "step": 12828
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.463464672681844e-05,
+      "loss": 14.3846,
+      "step": 12829
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.46307366958736e-05,
+      "loss": 15.6862,
+      "step": 12830
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.462682646602898e-05,
+      "loss": 16.9792,
+      "step": 12831
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.462291603731614e-05,
+      "loss": 16.6357,
+      "step": 12832
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.461900540976671e-05,
+      "loss": 15.3381,
+      "step": 12833
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.46150945834122e-05,
+      "loss": 17.061,
+      "step": 12834
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.461118355828425e-05,
+      "loss": 14.6444,
+      "step": 12835
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.460727233441442e-05,
+      "loss": 16.7541,
+      "step": 12836
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.460336091183429e-05,
+      "loss": 13.7594,
+      "step": 12837
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.459944929057543e-05,
+      "loss": 15.3183,
+      "step": 12838
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.459553747066947e-05,
+      "loss": 17.0473,
+      "step": 12839
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.459162545214795e-05,
+      "loss": 15.9846,
+      "step": 12840
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.458771323504251e-05,
+      "loss": 16.7751,
+      "step": 12841
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.45838008193847e-05,
+      "loss": 15.8936,
+      "step": 12842
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.457988820520615e-05,
+      "loss": 17.2636,
+      "step": 12843
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.457597539253842e-05,
+      "loss": 16.325,
+      "step": 12844
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.457206238141314e-05,
+      "loss": 15.7725,
+      "step": 12845
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.456814917186188e-05,
+      "loss": 16.1253,
+      "step": 12846
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.456423576391624e-05,
+      "loss": 15.0143,
+      "step": 12847
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.456032215760786e-05,
+      "loss": 16.1597,
+      "step": 12848
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.455640835296831e-05,
+      "loss": 14.6015,
+      "step": 12849
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.45524943500292e-05,
+      "loss": 16.211,
+      "step": 12850
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.454858014882215e-05,
+      "loss": 15.8464,
+      "step": 12851
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.454466574937874e-05,
+      "loss": 17.2331,
+      "step": 12852
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.45407511517306e-05,
+      "loss": 15.3548,
+      "step": 12853
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.453683635590935e-05,
+      "loss": 16.0914,
+      "step": 12854
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.45329213619466e-05,
+      "loss": 13.7173,
+      "step": 12855
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.452900616987394e-05,
+      "loss": 15.2868,
+      "step": 12856
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.452509077972301e-05,
+      "loss": 16.1802,
+      "step": 12857
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.452117519152542e-05,
+      "loss": 16.3669,
+      "step": 12858
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.451725940531277e-05,
+      "loss": 15.217,
+      "step": 12859
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.451334342111674e-05,
+      "loss": 17.3088,
+      "step": 12860
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.450942723896888e-05,
+      "loss": 15.9679,
+      "step": 12861
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.450551085890087e-05,
+      "loss": 16.5794,
+      "step": 12862
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.45015942809443e-05,
+      "loss": 15.5329,
+      "step": 12863
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.449767750513082e-05,
+      "loss": 15.6019,
+      "step": 12864
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.449376053149204e-05,
+      "loss": 16.3997,
+      "step": 12865
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.44898433600596e-05,
+      "loss": 16.39,
+      "step": 12866
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.448592599086515e-05,
+      "loss": 17.3388,
+      "step": 12867
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.448200842394029e-05,
+      "loss": 17.8868,
+      "step": 12868
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.44780906593167e-05,
+      "loss": 16.4232,
+      "step": 12869
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.447417269702597e-05,
+      "loss": 14.2701,
+      "step": 12870
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.447025453709975e-05,
+      "loss": 16.1603,
+      "step": 12871
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.44663361795697e-05,
+      "loss": 16.4037,
+      "step": 12872
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.446241762446744e-05,
+      "loss": 16.0018,
+      "step": 12873
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.445849887182464e-05,
+      "loss": 16.1655,
+      "step": 12874
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.445457992167291e-05,
+      "loss": 14.3878,
+      "step": 12875
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.445066077404393e-05,
+      "loss": 15.1433,
+      "step": 12876
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.444674142896932e-05,
+      "loss": 16.4037,
+      "step": 12877
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.444282188648075e-05,
+      "loss": 16.2748,
+      "step": 12878
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.443890214660987e-05,
+      "loss": 17.5189,
+      "step": 12879
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.44349822093883e-05,
+      "loss": 17.7021,
+      "step": 12880
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.443106207484776e-05,
+      "loss": 16.7014,
+      "step": 12881
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.442714174301984e-05,
+      "loss": 16.5529,
+      "step": 12882
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.442322121393624e-05,
+      "loss": 15.1057,
+      "step": 12883
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.44193004876286e-05,
+      "loss": 14.6635,
+      "step": 12884
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.441537956412858e-05,
+      "loss": 15.5389,
+      "step": 12885
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.441145844346787e-05,
+      "loss": 18.0103,
+      "step": 12886
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.440753712567808e-05,
+      "loss": 13.5961,
+      "step": 12887
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.440361561079093e-05,
+      "loss": 16.2664,
+      "step": 12888
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.439969389883805e-05,
+      "loss": 14.9513,
+      "step": 12889
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.439577198985114e-05,
+      "loss": 14.1136,
+      "step": 12890
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.439184988386183e-05,
+      "loss": 16.0819,
+      "step": 12891
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.438792758090185e-05,
+      "loss": 15.7804,
+      "step": 12892
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.438400508100282e-05,
+      "loss": 16.2558,
+      "step": 12893
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.438008238419643e-05,
+      "loss": 15.4233,
+      "step": 12894
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.437615949051438e-05,
+      "loss": 15.9084,
+      "step": 12895
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.437223639998833e-05,
+      "loss": 16.0028,
+      "step": 12896
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.436831311264995e-05,
+      "loss": 15.9485,
+      "step": 12897
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.436438962853093e-05,
+      "loss": 15.0075,
+      "step": 12898
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.436046594766297e-05,
+      "loss": 17.152,
+      "step": 12899
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.435654207007773e-05,
+      "loss": 15.1076,
+      "step": 12900
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.43526179958069e-05,
+      "loss": 17.0031,
+      "step": 12901
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.434869372488217e-05,
+      "loss": 15.8241,
+      "step": 12902
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.434476925733526e-05,
+      "loss": 14.1407,
+      "step": 12903
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.434084459319783e-05,
+      "loss": 16.6448,
+      "step": 12904
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.433691973250156e-05,
+      "loss": 14.9577,
+      "step": 12905
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.433299467527818e-05,
+      "loss": 17.3604,
+      "step": 12906
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.432906942155935e-05,
+      "loss": 15.1765,
+      "step": 12907
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.43251439713768e-05,
+      "loss": 16.1979,
+      "step": 12908
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.432121832476221e-05,
+      "loss": 15.6064,
+      "step": 12909
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.431729248174728e-05,
+      "loss": 15.0323,
+      "step": 12910
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.431336644236374e-05,
+      "loss": 15.9914,
+      "step": 12911
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.430944020664324e-05,
+      "loss": 15.3354,
+      "step": 12912
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.430551377461754e-05,
+      "loss": 15.0454,
+      "step": 12913
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 7.43015871463183e-05,
+      "loss": 16.5245,
+      "step": 12914
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.429766032177729e-05,
+      "loss": 15.3262,
+      "step": 12915
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.429373330102614e-05,
+      "loss": 17.6032,
+      "step": 12916
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.428980608409664e-05,
+      "loss": 16.5834,
+      "step": 12917
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.428587867102045e-05,
+      "loss": 14.9623,
+      "step": 12918
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.42819510618293e-05,
+      "loss": 14.9843,
+      "step": 12919
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.427802325655491e-05,
+      "loss": 15.739,
+      "step": 12920
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.427409525522901e-05,
+      "loss": 13.6897,
+      "step": 12921
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.427016705788329e-05,
+      "loss": 15.173,
+      "step": 12922
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.42662386645495e-05,
+      "loss": 17.8958,
+      "step": 12923
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.426231007525935e-05,
+      "loss": 16.5376,
+      "step": 12924
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.425838129004456e-05,
+      "loss": 14.611,
+      "step": 12925
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.425445230893686e-05,
+      "loss": 15.8983,
+      "step": 12926
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.425052313196799e-05,
+      "loss": 13.9717,
+      "step": 12927
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.424659375916965e-05,
+      "loss": 15.764,
+      "step": 12928
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.424266419057359e-05,
+      "loss": 14.4235,
+      "step": 12929
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.423873442621157e-05,
+      "loss": 16.5084,
+      "step": 12930
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.423480446611526e-05,
+      "loss": 16.9077,
+      "step": 12931
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.423087431031646e-05,
+      "loss": 14.379,
+      "step": 12932
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.422694395884686e-05,
+      "loss": 15.669,
+      "step": 12933
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.422301341173822e-05,
+      "loss": 17.3049,
+      "step": 12934
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.421908266902226e-05,
+      "loss": 15.4333,
+      "step": 12935
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.421515173073076e-05,
+      "loss": 16.9821,
+      "step": 12936
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.421122059689543e-05,
+      "loss": 18.2063,
+      "step": 12937
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.420728926754803e-05,
+      "loss": 14.9319,
+      "step": 12938
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.42033577427203e-05,
+      "loss": 16.631,
+      "step": 12939
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.4199426022444e-05,
+      "loss": 16.3239,
+      "step": 12940
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.419549410675087e-05,
+      "loss": 16.3169,
+      "step": 12941
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.419156199567265e-05,
+      "loss": 16.0554,
+      "step": 12942
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.418762968924111e-05,
+      "loss": 16.1891,
+      "step": 12943
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.418369718748802e-05,
+      "loss": 14.8782,
+      "step": 12944
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.417976449044508e-05,
+      "loss": 17.2269,
+      "step": 12945
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.417583159814409e-05,
+      "loss": 14.5465,
+      "step": 12946
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.417189851061681e-05,
+      "loss": 14.5231,
+      "step": 12947
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.4167965227895e-05,
+      "loss": 14.421,
+      "step": 12948
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.416403175001041e-05,
+      "loss": 15.5177,
+      "step": 12949
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.416009807699482e-05,
+      "loss": 15.6706,
+      "step": 12950
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.415616420887995e-05,
+      "loss": 15.1543,
+      "step": 12951
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.415223014569763e-05,
+      "loss": 15.5621,
+      "step": 12952
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.414829588747959e-05,
+      "loss": 15.3923,
+      "step": 12953
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.414436143425763e-05,
+      "loss": 14.7266,
+      "step": 12954
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.414042678606347e-05,
+      "loss": 16.3711,
+      "step": 12955
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.413649194292893e-05,
+      "loss": 16.3818,
+      "step": 12956
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.413255690488577e-05,
+      "loss": 17.0864,
+      "step": 12957
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.412862167196577e-05,
+      "loss": 16.9671,
+      "step": 12958
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.41246862442007e-05,
+      "loss": 16.7654,
+      "step": 12959
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.412075062162234e-05,
+      "loss": 15.3351,
+      "step": 12960
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.411681480426249e-05,
+      "loss": 16.5214,
+      "step": 12961
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.411287879215291e-05,
+      "loss": 17.1028,
+      "step": 12962
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.410894258532539e-05,
+      "loss": 15.3348,
+      "step": 12963
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.410500618381172e-05,
+      "loss": 16.1391,
+      "step": 12964
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.410106958764369e-05,
+      "loss": 16.4759,
+      "step": 12965
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.409713279685308e-05,
+      "loss": 13.8947,
+      "step": 12966
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.409319581147169e-05,
+      "loss": 15.5335,
+      "step": 12967
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.408925863153131e-05,
+      "loss": 15.7666,
+      "step": 12968
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.408532125706374e-05,
+      "loss": 16.3936,
+      "step": 12969
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.408138368810075e-05,
+      "loss": 14.263,
+      "step": 12970
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.407744592467416e-05,
+      "loss": 15.5865,
+      "step": 12971
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.407350796681577e-05,
+      "loss": 16.9605,
+      "step": 12972
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.406956981455737e-05,
+      "loss": 15.656,
+      "step": 12973
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.406563146793075e-05,
+      "loss": 15.3987,
+      "step": 12974
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.406169292696774e-05,
+      "loss": 14.727,
+      "step": 12975
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.405775419170014e-05,
+      "loss": 16.8487,
+      "step": 12976
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.405381526215973e-05,
+      "loss": 17.0699,
+      "step": 12977
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.404987613837835e-05,
+      "loss": 16.987,
+      "step": 12978
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.404593682038778e-05,
+      "loss": 17.5651,
+      "step": 12979
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.404199730821987e-05,
+      "loss": 15.6153,
+      "step": 12980
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.403805760190638e-05,
+      "loss": 15.3397,
+      "step": 12981
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.40341177014792e-05,
+      "loss": 15.1695,
+      "step": 12982
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.403017760697005e-05,
+      "loss": 14.2064,
+      "step": 12983
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.402623731841083e-05,
+      "loss": 14.3117,
+      "step": 12984
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.40222968358333e-05,
+      "loss": 16.6074,
+      "step": 12985
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.401835615926934e-05,
+      "loss": 15.3845,
+      "step": 12986
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.401441528875071e-05,
+      "loss": 14.292,
+      "step": 12987
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.401047422430927e-05,
+      "loss": 14.3805,
+      "step": 12988
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.400653296597682e-05,
+      "loss": 15.45,
+      "step": 12989
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.400259151378523e-05,
+      "loss": 15.4325,
+      "step": 12990
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.39986498677663e-05,
+      "loss": 16.6736,
+      "step": 12991
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.399470802795186e-05,
+      "loss": 15.1226,
+      "step": 12992
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.399076599437373e-05,
+      "loss": 14.7321,
+      "step": 12993
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.398682376706376e-05,
+      "loss": 14.8289,
+      "step": 12994
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.398288134605378e-05,
+      "loss": 15.3618,
+      "step": 12995
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.397893873137563e-05,
+      "loss": 17.6292,
+      "step": 12996
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.397499592306114e-05,
+      "loss": 16.3885,
+      "step": 12997
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.397105292114217e-05,
+      "loss": 15.0338,
+      "step": 12998
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.396710972565053e-05,
+      "loss": 16.4271,
+      "step": 12999
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.396316633661807e-05,
+      "loss": 17.8063,
+      "step": 13000
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.395922275407666e-05,
+      "loss": 15.8447,
+      "step": 13001
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.395527897805812e-05,
+      "loss": 16.2823,
+      "step": 13002
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.395133500859431e-05,
+      "loss": 15.1666,
+      "step": 13003
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.394739084571704e-05,
+      "loss": 17.1087,
+      "step": 13004
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.394344648945823e-05,
+      "loss": 16.5994,
+      "step": 13005
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.393950193984968e-05,
+      "loss": 17.1911,
+      "step": 13006
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.393555719692326e-05,
+      "loss": 16.8779,
+      "step": 13007
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.39316122607108e-05,
+      "loss": 15.2969,
+      "step": 13008
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.392766713124421e-05,
+      "loss": 15.4749,
+      "step": 13009
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.392372180855529e-05,
+      "loss": 18.2495,
+      "step": 13010
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.391977629267594e-05,
+      "loss": 15.9129,
+      "step": 13011
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.391583058363801e-05,
+      "loss": 15.2946,
+      "step": 13012
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.391188468147336e-05,
+      "loss": 15.7984,
+      "step": 13013
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.390793858621386e-05,
+      "loss": 17.4032,
+      "step": 13014
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.390399229789136e-05,
+      "loss": 13.3022,
+      "step": 13015
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.390004581653773e-05,
+      "loss": 18.5046,
+      "step": 13016
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.389609914218484e-05,
+      "loss": 16.0657,
+      "step": 13017
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.38921522748646e-05,
+      "loss": 14.8173,
+      "step": 13018
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.388820521460881e-05,
+      "loss": 14.5433,
+      "step": 13019
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.38842579614494e-05,
+      "loss": 16.194,
+      "step": 13020
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.388031051541823e-05,
+      "loss": 16.2118,
+      "step": 13021
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.387636287654717e-05,
+      "loss": 15.7621,
+      "step": 13022
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.387241504486809e-05,
+      "loss": 15.0721,
+      "step": 13023
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.386846702041291e-05,
+      "loss": 15.5898,
+      "step": 13024
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.386451880321345e-05,
+      "loss": 15.4735,
+      "step": 13025
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.386057039330166e-05,
+      "loss": 14.9848,
+      "step": 13026
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.385662179070935e-05,
+      "loss": 14.8981,
+      "step": 13027
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.385267299546847e-05,
+      "loss": 16.3783,
+      "step": 13028
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.384872400761089e-05,
+      "loss": 15.9793,
+      "step": 13029
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.384477482716848e-05,
+      "loss": 15.2604,
+      "step": 13030
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.384082545417315e-05,
+      "loss": 16.1354,
+      "step": 13031
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.383687588865679e-05,
+      "loss": 13.1035,
+      "step": 13032
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.383292613065127e-05,
+      "loss": 15.2816,
+      "step": 13033
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 7.382897618018853e-05,
+      "loss": 14.0889,
+      "step": 13034
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.382502603730044e-05,
+      "loss": 16.3738,
+      "step": 13035
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.38210757020189e-05,
+      "loss": 15.3841,
+      "step": 13036
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.381712517437579e-05,
+      "loss": 15.6295,
+      "step": 13037
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.381317445440305e-05,
+      "loss": 15.4163,
+      "step": 13038
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.380922354213256e-05,
+      "loss": 15.9366,
+      "step": 13039
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.380527243759623e-05,
+      "loss": 15.8807,
+      "step": 13040
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.380132114082596e-05,
+      "loss": 16.5815,
+      "step": 13041
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.379736965185368e-05,
+      "loss": 15.8853,
+      "step": 13042
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.379341797071128e-05,
+      "loss": 15.672,
+      "step": 13043
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.378946609743067e-05,
+      "loss": 17.6026,
+      "step": 13044
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.378551403204376e-05,
+      "loss": 15.6114,
+      "step": 13045
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.378156177458248e-05,
+      "loss": 15.1574,
+      "step": 13046
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.377760932507874e-05,
+      "loss": 14.7296,
+      "step": 13047
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.377365668356443e-05,
+      "loss": 15.2657,
+      "step": 13048
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.376970385007152e-05,
+      "loss": 15.334,
+      "step": 13049
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.376575082463187e-05,
+      "loss": 15.8508,
+      "step": 13050
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.376179760727747e-05,
+      "loss": 15.3614,
+      "step": 13051
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.375784419804019e-05,
+      "loss": 17.0003,
+      "step": 13052
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.375389059695196e-05,
+      "loss": 17.8369,
+      "step": 13053
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.374993680404474e-05,
+      "loss": 15.5091,
+      "step": 13054
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.374598281935042e-05,
+      "loss": 18.8846,
+      "step": 13055
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.374202864290096e-05,
+      "loss": 15.0091,
+      "step": 13056
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.373807427472827e-05,
+      "loss": 14.3168,
+      "step": 13057
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.373411971486429e-05,
+      "loss": 14.9793,
+      "step": 13058
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.373016496334095e-05,
+      "loss": 17.4266,
+      "step": 13059
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.37262100201902e-05,
+      "loss": 15.1917,
+      "step": 13060
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.372225488544395e-05,
+      "loss": 14.671,
+      "step": 13061
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.371829955913416e-05,
+      "loss": 14.7212,
+      "step": 13062
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.371434404129278e-05,
+      "loss": 16.4307,
+      "step": 13063
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.371038833195172e-05,
+      "loss": 15.1882,
+      "step": 13064
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.370643243114295e-05,
+      "loss": 16.2491,
+      "step": 13065
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.37024763388984e-05,
+      "loss": 16.8671,
+      "step": 13066
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.369852005525001e-05,
+      "loss": 15.6438,
+      "step": 13067
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.369456358022977e-05,
+      "loss": 16.4492,
+      "step": 13068
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.369060691386958e-05,
+      "loss": 16.9676,
+      "step": 13069
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.368665005620141e-05,
+      "loss": 15.0495,
+      "step": 13070
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.368269300725721e-05,
+      "loss": 14.4808,
+      "step": 13071
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.367873576706896e-05,
+      "loss": 15.4064,
+      "step": 13072
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.367477833566858e-05,
+      "loss": 15.5855,
+      "step": 13073
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.367082071308804e-05,
+      "loss": 15.9058,
+      "step": 13074
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.36668628993593e-05,
+      "loss": 16.2244,
+      "step": 13075
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.366290489451433e-05,
+      "loss": 15.8092,
+      "step": 13076
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.365894669858506e-05,
+      "loss": 15.403,
+      "step": 13077
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.365498831160349e-05,
+      "loss": 15.7681,
+      "step": 13078
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.365102973360157e-05,
+      "loss": 15.1806,
+      "step": 13079
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.364707096461127e-05,
+      "loss": 15.359,
+      "step": 13080
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.364311200466455e-05,
+      "loss": 15.6926,
+      "step": 13081
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.363915285379339e-05,
+      "loss": 13.1902,
+      "step": 13082
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.363519351202975e-05,
+      "loss": 16.4989,
+      "step": 13083
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.363123397940561e-05,
+      "loss": 14.9888,
+      "step": 13084
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.362727425595295e-05,
+      "loss": 15.036,
+      "step": 13085
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.362331434170373e-05,
+      "loss": 15.418,
+      "step": 13086
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.361935423668994e-05,
+      "loss": 14.256,
+      "step": 13087
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.361539394094357e-05,
+      "loss": 14.1901,
+      "step": 13088
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.361143345449656e-05,
+      "loss": 14.3403,
+      "step": 13089
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.360747277738094e-05,
+      "loss": 15.2527,
+      "step": 13090
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.360351190962865e-05,
+      "loss": 16.3952,
+      "step": 13091
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.35995508512717e-05,
+      "loss": 15.2719,
+      "step": 13092
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.359558960234209e-05,
+      "loss": 15.5341,
+      "step": 13093
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.359162816287177e-05,
+      "loss": 17.6732,
+      "step": 13094
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.358766653289277e-05,
+      "loss": 15.7928,
+      "step": 13095
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.358370471243705e-05,
+      "loss": 15.74,
+      "step": 13096
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.357974270153662e-05,
+      "loss": 15.4369,
+      "step": 13097
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.357578050022347e-05,
+      "loss": 14.6745,
+      "step": 13098
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.35718181085296e-05,
+      "loss": 14.8792,
+      "step": 13099
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.3567855526487e-05,
+      "loss": 15.2993,
+      "step": 13100
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.356389275412768e-05,
+      "loss": 15.2981,
+      "step": 13101
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.355992979148361e-05,
+      "loss": 16.4537,
+      "step": 13102
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.355596663858681e-05,
+      "loss": 16.2185,
+      "step": 13103
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.35520032954693e-05,
+      "loss": 17.1875,
+      "step": 13104
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.354803976216308e-05,
+      "loss": 16.132,
+      "step": 13105
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.354407603870013e-05,
+      "loss": 15.9276,
+      "step": 13106
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.354011212511249e-05,
+      "loss": 14.9982,
+      "step": 13107
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.353614802143216e-05,
+      "loss": 16.0923,
+      "step": 13108
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.353218372769113e-05,
+      "loss": 14.6956,
+      "step": 13109
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.352821924392144e-05,
+      "loss": 16.0771,
+      "step": 13110
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.35242545701551e-05,
+      "loss": 14.6441,
+      "step": 13111
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.352028970642412e-05,
+      "loss": 15.6874,
+      "step": 13112
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.35163246527605e-05,
+      "loss": 15.6679,
+      "step": 13113
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.35123594091963e-05,
+      "loss": 15.8962,
+      "step": 13114
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.350839397576351e-05,
+      "loss": 15.7692,
+      "step": 13115
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.350442835249416e-05,
+      "loss": 16.0846,
+      "step": 13116
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.350046253942027e-05,
+      "loss": 17.2893,
+      "step": 13117
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.349649653657386e-05,
+      "loss": 17.0183,
+      "step": 13118
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.349253034398698e-05,
+      "loss": 17.2766,
+      "step": 13119
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.348856396169164e-05,
+      "loss": 15.2803,
+      "step": 13120
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.348459738971986e-05,
+      "loss": 15.7114,
+      "step": 13121
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.348063062810369e-05,
+      "loss": 15.397,
+      "step": 13122
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.347666367687515e-05,
+      "loss": 17.2151,
+      "step": 13123
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.347269653606628e-05,
+      "loss": 14.1895,
+      "step": 13124
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.346872920570912e-05,
+      "loss": 19.4464,
+      "step": 13125
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.346476168583571e-05,
+      "loss": 15.25,
+      "step": 13126
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.346079397647807e-05,
+      "loss": 14.6497,
+      "step": 13127
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.345682607766826e-05,
+      "loss": 14.6552,
+      "step": 13128
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.34528579894383e-05,
+      "loss": 16.2845,
+      "step": 13129
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.344888971182027e-05,
+      "loss": 14.6539,
+      "step": 13130
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.344492124484617e-05,
+      "loss": 14.5017,
+      "step": 13131
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.344095258854807e-05,
+      "loss": 13.5318,
+      "step": 13132
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.343698374295802e-05,
+      "loss": 15.2381,
+      "step": 13133
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.343301470810808e-05,
+      "loss": 17.5755,
+      "step": 13134
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.342904548403027e-05,
+      "loss": 15.7585,
+      "step": 13135
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.342507607075666e-05,
+      "loss": 14.682,
+      "step": 13136
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.342110646831931e-05,
+      "loss": 15.0342,
+      "step": 13137
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.341713667675026e-05,
+      "loss": 16.4335,
+      "step": 13138
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.341316669608159e-05,
+      "loss": 15.3641,
+      "step": 13139
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.340919652634533e-05,
+      "loss": 16.2616,
+      "step": 13140
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.340522616757357e-05,
+      "loss": 15.4188,
+      "step": 13141
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.340125561979834e-05,
+      "loss": 15.3075,
+      "step": 13142
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.339728488305175e-05,
+      "loss": 16.3943,
+      "step": 13143
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.339331395736581e-05,
+      "loss": 15.017,
+      "step": 13144
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.338934284277262e-05,
+      "loss": 16.63,
+      "step": 13145
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.338537153930422e-05,
+      "loss": 15.5982,
+      "step": 13146
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.338140004699272e-05,
+      "loss": 14.5447,
+      "step": 13147
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.337742836587016e-05,
+      "loss": 16.8118,
+      "step": 13148
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.337345649596863e-05,
+      "loss": 14.3317,
+      "step": 13149
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.336948443732018e-05,
+      "loss": 15.281,
+      "step": 13150
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.336551218995691e-05,
+      "loss": 17.2136,
+      "step": 13151
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.33615397539109e-05,
+      "loss": 15.287,
+      "step": 13152
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.335756712921419e-05,
+      "loss": 15.5064,
+      "step": 13153
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 7.33535943158989e-05,
+      "loss": 15.7873,
+      "step": 13154
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.334962131399708e-05,
+      "loss": 14.6599,
+      "step": 13155
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.334564812354085e-05,
+      "loss": 16.3026,
+      "step": 13156
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.334167474456227e-05,
+      "loss": 14.6404,
+      "step": 13157
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.333770117709344e-05,
+      "loss": 14.7934,
+      "step": 13158
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.333372742116643e-05,
+      "loss": 16.2767,
+      "step": 13159
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.332975347681335e-05,
+      "loss": 15.117,
+      "step": 13160
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.332577934406626e-05,
+      "loss": 15.0949,
+      "step": 13161
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.33218050229573e-05,
+      "loss": 18.6181,
+      "step": 13162
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.331783051351849e-05,
+      "loss": 14.8719,
+      "step": 13163
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.3313855815782e-05,
+      "loss": 16.3548,
+      "step": 13164
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.33098809297799e-05,
+      "loss": 15.2699,
+      "step": 13165
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.330590585554428e-05,
+      "loss": 14.9905,
+      "step": 13166
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.330193059310723e-05,
+      "loss": 17.0203,
+      "step": 13167
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.329795514250088e-05,
+      "loss": 14.4885,
+      "step": 13168
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.329397950375731e-05,
+      "loss": 17.9389,
+      "step": 13169
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.329000367690864e-05,
+      "loss": 16.4421,
+      "step": 13170
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.328602766198698e-05,
+      "loss": 16.6692,
+      "step": 13171
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.328205145902441e-05,
+      "loss": 14.9672,
+      "step": 13172
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.327807506805306e-05,
+      "loss": 16.9543,
+      "step": 13173
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.327409848910505e-05,
+      "loss": 15.6774,
+      "step": 13174
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.327012172221245e-05,
+      "loss": 15.166,
+      "step": 13175
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.326614476740741e-05,
+      "loss": 15.8686,
+      "step": 13176
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.326216762472205e-05,
+      "loss": 14.6933,
+      "step": 13177
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.325819029418847e-05,
+      "loss": 15.9365,
+      "step": 13178
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.325421277583876e-05,
+      "loss": 15.4329,
+      "step": 13179
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.325023506970512e-05,
+      "loss": 15.8082,
+      "step": 13180
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.324625717581959e-05,
+      "loss": 14.649,
+      "step": 13181
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.324227909421433e-05,
+      "loss": 15.3521,
+      "step": 13182
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.323830082492146e-05,
+      "loss": 15.9396,
+      "step": 13183
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.323432236797309e-05,
+      "loss": 15.5823,
+      "step": 13184
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.323034372340139e-05,
+      "loss": 15.7519,
+      "step": 13185
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.322636489123843e-05,
+      "loss": 14.1468,
+      "step": 13186
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.322238587151639e-05,
+      "loss": 15.7108,
+      "step": 13187
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.321840666426737e-05,
+      "loss": 16.4018,
+      "step": 13188
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.321442726952351e-05,
+      "loss": 17.5268,
+      "step": 13189
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.321044768731696e-05,
+      "loss": 15.4915,
+      "step": 13190
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.320646791767984e-05,
+      "loss": 16.7251,
+      "step": 13191
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.320248796064428e-05,
+      "loss": 15.7031,
+      "step": 13192
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.319850781624245e-05,
+      "loss": 16.5842,
+      "step": 13193
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.319452748450646e-05,
+      "loss": 15.7634,
+      "step": 13194
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.319054696546848e-05,
+      "loss": 15.6823,
+      "step": 13195
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.318656625916063e-05,
+      "loss": 14.2822,
+      "step": 13196
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.318258536561506e-05,
+      "loss": 16.2574,
+      "step": 13197
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.317860428486391e-05,
+      "loss": 14.339,
+      "step": 13198
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.317462301693936e-05,
+      "loss": 16.0311,
+      "step": 13199
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.317064156187354e-05,
+      "loss": 15.7312,
+      "step": 13200
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.316665991969858e-05,
+      "loss": 15.2561,
+      "step": 13201
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.316267809044667e-05,
+      "loss": 16.0894,
+      "step": 13202
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.315869607414993e-05,
+      "loss": 16.9453,
+      "step": 13203
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.315471387084056e-05,
+      "loss": 15.4447,
+      "step": 13204
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.315073148055067e-05,
+      "loss": 16.8378,
+      "step": 13205
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.314674890331245e-05,
+      "loss": 17.1224,
+      "step": 13206
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.314276613915804e-05,
+      "loss": 17.438,
+      "step": 13207
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.313878318811963e-05,
+      "loss": 16.3313,
+      "step": 13208
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.313480005022933e-05,
+      "loss": 16.8482,
+      "step": 13209
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.313081672551937e-05,
+      "loss": 16.1235,
+      "step": 13210
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.312683321402187e-05,
+      "loss": 15.8349,
+      "step": 13211
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.312284951576902e-05,
+      "loss": 15.9821,
+      "step": 13212
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.311886563079298e-05,
+      "loss": 17.591,
+      "step": 13213
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.311488155912594e-05,
+      "loss": 17.1245,
+      "step": 13214
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.311089730080004e-05,
+      "loss": 15.3009,
+      "step": 13215
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.310691285584747e-05,
+      "loss": 16.8324,
+      "step": 13216
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.310292822430042e-05,
+      "loss": 14.7962,
+      "step": 13217
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.309894340619106e-05,
+      "loss": 16.4152,
+      "step": 13218
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.309495840155154e-05,
+      "loss": 14.563,
+      "step": 13219
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.309097321041406e-05,
+      "loss": 16.6616,
+      "step": 13220
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.308698783281081e-05,
+      "loss": 16.3326,
+      "step": 13221
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.308300226877399e-05,
+      "loss": 16.8781,
+      "step": 13222
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.307901651833573e-05,
+      "loss": 15.3012,
+      "step": 13223
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.307503058152826e-05,
+      "loss": 13.9901,
+      "step": 13224
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.307104445838376e-05,
+      "loss": 15.8738,
+      "step": 13225
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.30670581489344e-05,
+      "loss": 15.2568,
+      "step": 13226
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.306307165321238e-05,
+      "loss": 16.7123,
+      "step": 13227
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.305908497124992e-05,
+      "loss": 15.7866,
+      "step": 13228
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.305509810307917e-05,
+      "loss": 18.2484,
+      "step": 13229
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.305111104873235e-05,
+      "loss": 17.1705,
+      "step": 13230
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.304712380824165e-05,
+      "loss": 16.6739,
+      "step": 13231
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.304313638163926e-05,
+      "loss": 16.9348,
+      "step": 13232
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.30391487689574e-05,
+      "loss": 13.3501,
+      "step": 13233
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.303516097022824e-05,
+      "loss": 14.5681,
+      "step": 13234
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.303117298548404e-05,
+      "loss": 16.0789,
+      "step": 13235
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.302718481475693e-05,
+      "loss": 15.2233,
+      "step": 13236
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.302319645807917e-05,
+      "loss": 14.58,
+      "step": 13237
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.301920791548294e-05,
+      "loss": 15.6246,
+      "step": 13238
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.301521918700046e-05,
+      "loss": 14.4858,
+      "step": 13239
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.301123027266393e-05,
+      "loss": 16.1596,
+      "step": 13240
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.300724117250559e-05,
+      "loss": 14.9017,
+      "step": 13241
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.300325188655761e-05,
+      "loss": 15.3756,
+      "step": 13242
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.299926241485224e-05,
+      "loss": 15.6604,
+      "step": 13243
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.299527275742167e-05,
+      "loss": 16.1335,
+      "step": 13244
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.299128291429813e-05,
+      "loss": 15.7515,
+      "step": 13245
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.298729288551385e-05,
+      "loss": 16.3167,
+      "step": 13246
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.298330267110103e-05,
+      "loss": 13.6597,
+      "step": 13247
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.297931227109193e-05,
+      "loss": 16.4419,
+      "step": 13248
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.297532168551872e-05,
+      "loss": 17.7122,
+      "step": 13249
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.297133091441364e-05,
+      "loss": 16.7811,
+      "step": 13250
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.296733995780894e-05,
+      "loss": 14.068,
+      "step": 13251
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.296334881573685e-05,
+      "loss": 14.8771,
+      "step": 13252
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.295935748822956e-05,
+      "loss": 17.0715,
+      "step": 13253
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.295536597531934e-05,
+      "loss": 16.0675,
+      "step": 13254
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.29513742770384e-05,
+      "loss": 14.6638,
+      "step": 13255
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.294738239341899e-05,
+      "loss": 14.5619,
+      "step": 13256
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.294339032449332e-05,
+      "loss": 13.8393,
+      "step": 13257
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.293939807029366e-05,
+      "loss": 17.2635,
+      "step": 13258
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.293540563085222e-05,
+      "loss": 15.0543,
+      "step": 13259
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.293141300620127e-05,
+      "loss": 14.4854,
+      "step": 13260
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.292742019637303e-05,
+      "loss": 16.7557,
+      "step": 13261
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.292342720139975e-05,
+      "loss": 15.0971,
+      "step": 13262
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.291943402131366e-05,
+      "loss": 16.1486,
+      "step": 13263
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.291544065614703e-05,
+      "loss": 15.9239,
+      "step": 13264
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.291144710593208e-05,
+      "loss": 15.8915,
+      "step": 13265
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.29074533707011e-05,
+      "loss": 17.0791,
+      "step": 13266
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.290345945048629e-05,
+      "loss": 19.0953,
+      "step": 13267
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.289946534531994e-05,
+      "loss": 15.6844,
+      "step": 13268
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.289547105523429e-05,
+      "loss": 15.8153,
+      "step": 13269
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.289147658026159e-05,
+      "loss": 17.061,
+      "step": 13270
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.288748192043411e-05,
+      "loss": 16.6942,
+      "step": 13271
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.288348707578408e-05,
+      "loss": 15.3934,
+      "step": 13272
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.28794920463438e-05,
+      "loss": 15.85,
+      "step": 13273
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.28754968321455e-05,
+      "loss": 15.323,
+      "step": 13274
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.287150143322147e-05,
+      "loss": 16.0676,
+      "step": 13275
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.286750584960394e-05,
+      "loss": 13.4899,
+      "step": 13276
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.28635100813252e-05,
+      "loss": 17.5406,
+      "step": 13277
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.28595141284175e-05,
+      "loss": 16.4408,
+      "step": 13278
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.285551799091311e-05,
+      "loss": 15.6719,
+      "step": 13279
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.285152166884432e-05,
+      "loss": 15.3877,
+      "step": 13280
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.284752516224339e-05,
+      "loss": 15.365,
+      "step": 13281
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.284352847114259e-05,
+      "loss": 17.2867,
+      "step": 13282
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.28395315955742e-05,
+      "loss": 14.8694,
+      "step": 13283
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.283553453557047e-05,
+      "loss": 16.4458,
+      "step": 13284
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.283153729116372e-05,
+      "loss": 16.1448,
+      "step": 13285
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.28275398623862e-05,
+      "loss": 16.8943,
+      "step": 13286
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.282354224927019e-05,
+      "loss": 17.5048,
+      "step": 13287
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.281954445184798e-05,
+      "loss": 17.6909,
+      "step": 13288
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.281554647015187e-05,
+      "loss": 14.7592,
+      "step": 13289
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.281154830421413e-05,
+      "loss": 13.9248,
+      "step": 13290
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.280754995406702e-05,
+      "loss": 16.6109,
+      "step": 13291
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.280355141974286e-05,
+      "loss": 16.2633,
+      "step": 13292
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.279955270127394e-05,
+      "loss": 15.714,
+      "step": 13293
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.279555379869254e-05,
+      "loss": 16.5348,
+      "step": 13294
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.279155471203095e-05,
+      "loss": 15.6486,
+      "step": 13295
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.278755544132147e-05,
+      "loss": 16.698,
+      "step": 13296
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.278355598659639e-05,
+      "loss": 15.119,
+      "step": 13297
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.277955634788802e-05,
+      "loss": 14.5965,
+      "step": 13298
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.277555652522863e-05,
+      "loss": 14.5992,
+      "step": 13299
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.277155651865055e-05,
+      "loss": 15.2638,
+      "step": 13300
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.276755632818606e-05,
+      "loss": 15.345,
+      "step": 13301
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.276355595386748e-05,
+      "loss": 16.9645,
+      "step": 13302
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.27595553957271e-05,
+      "loss": 14.2954,
+      "step": 13303
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.275555465379723e-05,
+      "loss": 16.5149,
+      "step": 13304
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.275155372811018e-05,
+      "loss": 17.066,
+      "step": 13305
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.274755261869825e-05,
+      "loss": 14.4243,
+      "step": 13306
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.274355132559376e-05,
+      "loss": 16.0417,
+      "step": 13307
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.273954984882902e-05,
+      "loss": 17.2847,
+      "step": 13308
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.273554818843634e-05,
+      "loss": 15.0541,
+      "step": 13309
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.273154634444804e-05,
+      "loss": 16.6765,
+      "step": 13310
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.272754431689643e-05,
+      "loss": 16.3211,
+      "step": 13311
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.272354210581383e-05,
+      "loss": 16.137,
+      "step": 13312
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.271953971123254e-05,
+      "loss": 17.7671,
+      "step": 13313
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.271553713318492e-05,
+      "loss": 14.5234,
+      "step": 13314
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.271153437170326e-05,
+      "loss": 14.757,
+      "step": 13315
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.27075314268199e-05,
+      "loss": 15.9509,
+      "step": 13316
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.270352829856715e-05,
+      "loss": 15.8969,
+      "step": 13317
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.269952498697734e-05,
+      "loss": 16.8757,
+      "step": 13318
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.269552149208281e-05,
+      "loss": 17.4535,
+      "step": 13319
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.269151781391589e-05,
+      "loss": 16.1066,
+      "step": 13320
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.268751395250889e-05,
+      "loss": 14.7069,
+      "step": 13321
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.268350990789416e-05,
+      "loss": 15.8646,
+      "step": 13322
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.267950568010402e-05,
+      "loss": 15.5196,
+      "step": 13323
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.267550126917082e-05,
+      "loss": 15.7726,
+      "step": 13324
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.26714966751269e-05,
+      "loss": 17.3647,
+      "step": 13325
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.266749189800457e-05,
+      "loss": 13.7064,
+      "step": 13326
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.26634869378362e-05,
+      "loss": 16.5164,
+      "step": 13327
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.265948179465411e-05,
+      "loss": 14.9689,
+      "step": 13328
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.265547646849065e-05,
+      "loss": 16.9649,
+      "step": 13329
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.265147095937818e-05,
+      "loss": 14.8539,
+      "step": 13330
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.264746526734902e-05,
+      "loss": 13.1999,
+      "step": 13331
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.264345939243553e-05,
+      "loss": 15.9624,
+      "step": 13332
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.263945333467006e-05,
+      "loss": 15.9429,
+      "step": 13333
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.263544709408495e-05,
+      "loss": 16.2676,
+      "step": 13334
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.263144067071257e-05,
+      "loss": 15.4185,
+      "step": 13335
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.262743406458525e-05,
+      "loss": 15.5406,
+      "step": 13336
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.262342727573535e-05,
+      "loss": 15.7408,
+      "step": 13337
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.261942030419525e-05,
+      "loss": 13.4438,
+      "step": 13338
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.261541314999728e-05,
+      "loss": 19.1229,
+      "step": 13339
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.261140581317382e-05,
+      "loss": 16.0163,
+      "step": 13340
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.26073982937572e-05,
+      "loss": 14.3878,
+      "step": 13341
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.260339059177981e-05,
+      "loss": 16.7445,
+      "step": 13342
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.2599382707274e-05,
+      "loss": 15.6137,
+      "step": 13343
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.259537464027214e-05,
+      "loss": 14.8115,
+      "step": 13344
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.259136639080658e-05,
+      "loss": 15.3526,
+      "step": 13345
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.258735795890972e-05,
+      "loss": 18.3413,
+      "step": 13346
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.25833493446139e-05,
+      "loss": 17.2172,
+      "step": 13347
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.257934054795152e-05,
+      "loss": 15.7496,
+      "step": 13348
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.257533156895491e-05,
+      "loss": 15.9419,
+      "step": 13349
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.257132240765649e-05,
+      "loss": 14.8528,
+      "step": 13350
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.25673130640886e-05,
+      "loss": 15.0761,
+      "step": 13351
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.256330353828362e-05,
+      "loss": 16.4198,
+      "step": 13352
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.255929383027395e-05,
+      "loss": 14.904,
+      "step": 13353
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.255528394009195e-05,
+      "loss": 16.6296,
+      "step": 13354
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.255127386776999e-05,
+      "loss": 16.5319,
+      "step": 13355
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.25472636133405e-05,
+      "loss": 15.5428,
+      "step": 13356
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.254325317683583e-05,
+      "loss": 14.9802,
+      "step": 13357
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.253924255828836e-05,
+      "loss": 15.4949,
+      "step": 13358
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.253523175773049e-05,
+      "loss": 15.7142,
+      "step": 13359
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.25312207751946e-05,
+      "loss": 16.5133,
+      "step": 13360
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.252720961071308e-05,
+      "loss": 16.1013,
+      "step": 13361
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.252319826431833e-05,
+      "loss": 15.2822,
+      "step": 13362
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.251918673604274e-05,
+      "loss": 16.8824,
+      "step": 13363
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.251517502591869e-05,
+      "loss": 16.5268,
+      "step": 13364
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.251116313397861e-05,
+      "loss": 15.7276,
+      "step": 13365
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.250715106025486e-05,
+      "loss": 15.9504,
+      "step": 13366
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.250313880477985e-05,
+      "loss": 14.6374,
+      "step": 13367
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.249912636758598e-05,
+      "loss": 16.0262,
+      "step": 13368
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.249511374870566e-05,
+      "loss": 15.574,
+      "step": 13369
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.249110094817128e-05,
+      "loss": 14.5,
+      "step": 13370
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.248708796601528e-05,
+      "loss": 15.0656,
+      "step": 13371
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.248307480227001e-05,
+      "loss": 17.3118,
+      "step": 13372
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.247906145696793e-05,
+      "loss": 14.9302,
+      "step": 13373
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.24750479301414e-05,
+      "loss": 16.2365,
+      "step": 13374
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.247103422182286e-05,
+      "loss": 16.5412,
+      "step": 13375
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.24670203320447e-05,
+      "loss": 14.9096,
+      "step": 13376
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.246300626083938e-05,
+      "loss": 14.4996,
+      "step": 13377
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.245899200823927e-05,
+      "loss": 16.8164,
+      "step": 13378
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.245497757427681e-05,
+      "loss": 17.7446,
+      "step": 13379
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.245096295898439e-05,
+      "loss": 15.7615,
+      "step": 13380
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.244694816239447e-05,
+      "loss": 16.4072,
+      "step": 13381
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.244293318453943e-05,
+      "loss": 16.3088,
+      "step": 13382
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.24389180254517e-05,
+      "loss": 15.0239,
+      "step": 13383
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.243490268516374e-05,
+      "loss": 16.8953,
+      "step": 13384
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.243088716370794e-05,
+      "loss": 16.3201,
+      "step": 13385
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.242687146111673e-05,
+      "loss": 16.2559,
+      "step": 13386
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.242285557742254e-05,
+      "loss": 15.5634,
+      "step": 13387
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.24188395126578e-05,
+      "loss": 16.9667,
+      "step": 13388
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.241482326685494e-05,
+      "loss": 13.4769,
+      "step": 13389
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.241080684004641e-05,
+      "loss": 15.0058,
+      "step": 13390
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.240679023226462e-05,
+      "loss": 15.1778,
+      "step": 13391
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.240277344354202e-05,
+      "loss": 14.3094,
+      "step": 13392
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.239875647391103e-05,
+      "loss": 16.234,
+      "step": 13393
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.23947393234041e-05,
+      "loss": 16.2146,
+      "step": 13394
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.239072199205367e-05,
+      "loss": 16.6524,
+      "step": 13395
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.238670447989218e-05,
+      "loss": 14.5316,
+      "step": 13396
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.238268678695206e-05,
+      "loss": 15.5113,
+      "step": 13397
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.237866891326579e-05,
+      "loss": 16.5514,
+      "step": 13398
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.237465085886577e-05,
+      "loss": 16.5799,
+      "step": 13399
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.237063262378447e-05,
+      "loss": 14.62,
+      "step": 13400
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.236661420805435e-05,
+      "loss": 14.0476,
+      "step": 13401
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.236259561170783e-05,
+      "loss": 15.4113,
+      "step": 13402
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.235857683477739e-05,
+      "loss": 16.898,
+      "step": 13403
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.235455787729546e-05,
+      "loss": 15.454,
+      "step": 13404
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.23505387392945e-05,
+      "loss": 14.6926,
+      "step": 13405
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.234651942080697e-05,
+      "loss": 15.6423,
+      "step": 13406
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.234249992186533e-05,
+      "loss": 16.497,
+      "step": 13407
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.233848024250204e-05,
+      "loss": 16.6253,
+      "step": 13408
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.233446038274956e-05,
+      "loss": 16.5899,
+      "step": 13409
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.233044034264034e-05,
+      "loss": 14.4647,
+      "step": 13410
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.232642012220684e-05,
+      "loss": 15.3161,
+      "step": 13411
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.232239972148154e-05,
+      "loss": 16.2025,
+      "step": 13412
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.23183791404969e-05,
+      "loss": 15.4478,
+      "step": 13413
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.231435837928538e-05,
+      "loss": 17.6455,
+      "step": 13414
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.231033743787947e-05,
+      "loss": 16.2768,
+      "step": 13415
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.230631631631161e-05,
+      "loss": 15.744,
+      "step": 13416
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.230229501461431e-05,
+      "loss": 16.7035,
+      "step": 13417
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.229827353281999e-05,
+      "loss": 16.0619,
+      "step": 13418
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.229425187096118e-05,
+      "loss": 16.2914,
+      "step": 13419
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.229023002907031e-05,
+      "loss": 15.1453,
+      "step": 13420
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.228620800717991e-05,
+      "loss": 16.5119,
+      "step": 13421
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.22821858053224e-05,
+      "loss": 14.3011,
+      "step": 13422
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.22781634235303e-05,
+      "loss": 15.2096,
+      "step": 13423
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.227414086183608e-05,
+      "loss": 16.5801,
+      "step": 13424
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.227011812027221e-05,
+      "loss": 17.4282,
+      "step": 13425
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.226609519887118e-05,
+      "loss": 14.9388,
+      "step": 13426
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.226207209766552e-05,
+      "loss": 13.8783,
+      "step": 13427
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.225804881668766e-05,
+      "loss": 17.5866,
+      "step": 13428
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.22540253559701e-05,
+      "loss": 14.4214,
+      "step": 13429
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.225000171554534e-05,
+      "loss": 16.6193,
+      "step": 13430
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.224597789544588e-05,
+      "loss": 15.116,
+      "step": 13431
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.224195389570422e-05,
+      "loss": 14.6515,
+      "step": 13432
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.223792971635282e-05,
+      "loss": 15.7025,
+      "step": 13433
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.223390535742421e-05,
+      "loss": 16.0665,
+      "step": 13434
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.222988081895087e-05,
+      "loss": 15.1251,
+      "step": 13435
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.222585610096531e-05,
+      "loss": 16.4518,
+      "step": 13436
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.222183120350003e-05,
+      "loss": 14.8701,
+      "step": 13437
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.221780612658752e-05,
+      "loss": 15.554,
+      "step": 13438
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.221378087026028e-05,
+      "loss": 17.1213,
+      "step": 13439
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.220975543455084e-05,
+      "loss": 16.2504,
+      "step": 13440
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.220572981949168e-05,
+      "loss": 15.6547,
+      "step": 13441
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.220170402511534e-05,
+      "loss": 16.0525,
+      "step": 13442
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.21976780514543e-05,
+      "loss": 15.1339,
+      "step": 13443
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.219365189854108e-05,
+      "loss": 16.7185,
+      "step": 13444
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.218962556640819e-05,
+      "loss": 14.864,
+      "step": 13445
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.218559905508814e-05,
+      "loss": 14.6673,
+      "step": 13446
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.218157236461346e-05,
+      "loss": 16.6427,
+      "step": 13447
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.217754549501665e-05,
+      "loss": 16.368,
+      "step": 13448
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.217351844633025e-05,
+      "loss": 16.8846,
+      "step": 13449
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.216949121858676e-05,
+      "loss": 18.8032,
+      "step": 13450
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.21654638118187e-05,
+      "loss": 16.6638,
+      "step": 13451
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.21614362260586e-05,
+      "loss": 16.4916,
+      "step": 13452
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.215740846133898e-05,
+      "loss": 17.1162,
+      "step": 13453
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.215338051769237e-05,
+      "loss": 14.7834,
+      "step": 13454
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.21493523951513e-05,
+      "loss": 15.9253,
+      "step": 13455
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.214532409374829e-05,
+      "loss": 15.4515,
+      "step": 13456
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.214129561351588e-05,
+      "loss": 16.1004,
+      "step": 13457
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.213726695448658e-05,
+      "loss": 15.7486,
+      "step": 13458
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.213323811669294e-05,
+      "loss": 16.8054,
+      "step": 13459
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.212920910016749e-05,
+      "loss": 16.0223,
+      "step": 13460
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.212517990494278e-05,
+      "loss": 16.4607,
+      "step": 13461
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.212115053105132e-05,
+      "loss": 18.1608,
+      "step": 13462
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.211712097852566e-05,
+      "loss": 13.7348,
+      "step": 13463
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.211309124739834e-05,
+      "loss": 15.9329,
+      "step": 13464
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.210906133770191e-05,
+      "loss": 14.2329,
+      "step": 13465
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.210503124946889e-05,
+      "loss": 16.9029,
+      "step": 13466
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.210100098273185e-05,
+      "loss": 16.4432,
+      "step": 13467
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.209697053752331e-05,
+      "loss": 14.9663,
+      "step": 13468
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.209293991387584e-05,
+      "loss": 14.4337,
+      "step": 13469
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.208890911182197e-05,
+      "loss": 15.7476,
+      "step": 13470
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.208487813139428e-05,
+      "loss": 14.7539,
+      "step": 13471
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.208084697262528e-05,
+      "loss": 16.3913,
+      "step": 13472
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.207681563554755e-05,
+      "loss": 15.8845,
+      "step": 13473
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.207278412019363e-05,
+      "loss": 17.7353,
+      "step": 13474
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.206875242659607e-05,
+      "loss": 14.4007,
+      "step": 13475
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.206472055478747e-05,
+      "loss": 15.6949,
+      "step": 13476
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.206068850480034e-05,
+      "loss": 16.0409,
+      "step": 13477
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.205665627666726e-05,
+      "loss": 15.9428,
+      "step": 13478
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.205262387042079e-05,
+      "loss": 14.1371,
+      "step": 13479
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.204859128609349e-05,
+      "loss": 14.678,
+      "step": 13480
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.204455852371793e-05,
+      "loss": 14.4561,
+      "step": 13481
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.204052558332668e-05,
+      "loss": 15.6027,
+      "step": 13482
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.203649246495228e-05,
+      "loss": 15.8922,
+      "step": 13483
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.203245916862732e-05,
+      "loss": 15.8195,
+      "step": 13484
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.202842569438437e-05,
+      "loss": 17.5283,
+      "step": 13485
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.202439204225601e-05,
+      "loss": 14.9961,
+      "step": 13486
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.202035821227479e-05,
+      "loss": 15.9224,
+      "step": 13487
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.20163242044733e-05,
+      "loss": 15.3831,
+      "step": 13488
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.20122900188841e-05,
+      "loss": 14.7934,
+      "step": 13489
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.200825565553978e-05,
+      "loss": 16.4919,
+      "step": 13490
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.200422111447292e-05,
+      "loss": 16.2321,
+      "step": 13491
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.20001863957161e-05,
+      "loss": 15.0346,
+      "step": 13492
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.199615149930191e-05,
+      "loss": 14.6547,
+      "step": 13493
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.199211642526289e-05,
+      "loss": 14.4518,
+      "step": 13494
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.198808117363168e-05,
+      "loss": 15.4136,
+      "step": 13495
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.198404574444082e-05,
+      "loss": 16.4777,
+      "step": 13496
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.198001013772294e-05,
+      "loss": 15.5621,
+      "step": 13497
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.19759743535106e-05,
+      "loss": 14.4452,
+      "step": 13498
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.19719383918364e-05,
+      "loss": 16.5298,
+      "step": 13499
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.19679022527329e-05,
+      "loss": 14.4665,
+      "step": 13500
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.196386593623277e-05,
+      "loss": 18.2754,
+      "step": 13501
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.195982944236851e-05,
+      "loss": 15.7598,
+      "step": 13502
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.195579277117279e-05,
+      "loss": 17.7475,
+      "step": 13503
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.195175592267817e-05,
+      "loss": 15.4634,
+      "step": 13504
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.194771889691728e-05,
+      "loss": 15.8839,
+      "step": 13505
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.194368169392266e-05,
+      "loss": 15.2128,
+      "step": 13506
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.193964431372695e-05,
+      "loss": 15.6568,
+      "step": 13507
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.193560675636277e-05,
+      "loss": 17.1779,
+      "step": 13508
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.193156902186271e-05,
+      "loss": 15.2809,
+      "step": 13509
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.192753111025936e-05,
+      "loss": 15.6762,
+      "step": 13510
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.192349302158534e-05,
+      "loss": 16.3103,
+      "step": 13511
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.191945475587326e-05,
+      "loss": 15.8731,
+      "step": 13512
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.191541631315573e-05,
+      "loss": 14.4072,
+      "step": 13513
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.191137769346535e-05,
+      "loss": 15.1822,
+      "step": 13514
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.190733889683476e-05,
+      "loss": 16.2892,
+      "step": 13515
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.190329992329655e-05,
+      "loss": 15.9807,
+      "step": 13516
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.189926077288333e-05,
+      "loss": 16.23,
+      "step": 13517
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.189522144562773e-05,
+      "loss": 15.2934,
+      "step": 13518
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.18911819415624e-05,
+      "loss": 15.7086,
+      "step": 13519
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.188714226071989e-05,
+      "loss": 15.3219,
+      "step": 13520
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.188310240313286e-05,
+      "loss": 15.9855,
+      "step": 13521
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.187906236883395e-05,
+      "loss": 15.5039,
+      "step": 13522
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.187502215785576e-05,
+      "loss": 15.7527,
+      "step": 13523
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.187098177023092e-05,
+      "loss": 15.2219,
+      "step": 13524
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.186694120599208e-05,
+      "loss": 15.3124,
+      "step": 13525
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.186290046517184e-05,
+      "loss": 16.984,
+      "step": 13526
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.185885954780282e-05,
+      "loss": 14.5031,
+      "step": 13527
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.185481845391771e-05,
+      "loss": 15.059,
+      "step": 13528
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.185077718354907e-05,
+      "loss": 15.1864,
+      "step": 13529
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.184673573672957e-05,
+      "loss": 15.3443,
+      "step": 13530
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.184269411349184e-05,
+      "loss": 13.9862,
+      "step": 13531
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.183865231386854e-05,
+      "loss": 15.6343,
+      "step": 13532
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.183461033789228e-05,
+      "loss": 15.6754,
+      "step": 13533
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.183056818559571e-05,
+      "loss": 14.9724,
+      "step": 13534
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.182652585701145e-05,
+      "loss": 14.4973,
+      "step": 13535
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.182248335217218e-05,
+      "loss": 17.7611,
+      "step": 13536
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.181844067111053e-05,
+      "loss": 16.6696,
+      "step": 13537
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.181439781385914e-05,
+      "loss": 14.9689,
+      "step": 13538
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.181035478045066e-05,
+      "loss": 15.4028,
+      "step": 13539
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.180631157091774e-05,
+      "loss": 15.3794,
+      "step": 13540
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.180226818529303e-05,
+      "loss": 16.2999,
+      "step": 13541
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.179822462360917e-05,
+      "loss": 15.756,
+      "step": 13542
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.179418088589884e-05,
+      "loss": 16.602,
+      "step": 13543
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.179013697219468e-05,
+      "loss": 16.4238,
+      "step": 13544
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.178609288252934e-05,
+      "loss": 16.5741,
+      "step": 13545
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.178204861693545e-05,
+      "loss": 15.4179,
+      "step": 13546
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.177800417544575e-05,
+      "loss": 17.8103,
+      "step": 13547
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.17739595580928e-05,
+      "loss": 15.3502,
+      "step": 13548
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.176991476490935e-05,
+      "loss": 16.7386,
+      "step": 13549
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.176586979592799e-05,
+      "loss": 14.7567,
+      "step": 13550
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.176182465118145e-05,
+      "loss": 15.2735,
+      "step": 13551
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.175777933070233e-05,
+      "loss": 17.6036,
+      "step": 13552
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.175373383452336e-05,
+      "loss": 15.1996,
+      "step": 13553
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.174968816267716e-05,
+      "loss": 16.6991,
+      "step": 13554
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.174564231519643e-05,
+      "loss": 15.7787,
+      "step": 13555
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.174159629211382e-05,
+      "loss": 15.4548,
+      "step": 13556
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.173755009346203e-05,
+      "loss": 14.9191,
+      "step": 13557
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.173350371927369e-05,
+      "loss": 15.6107,
+      "step": 13558
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.172945716958152e-05,
+      "loss": 15.6539,
+      "step": 13559
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.172541044441817e-05,
+      "loss": 15.1784,
+      "step": 13560
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.172136354381633e-05,
+      "loss": 15.4548,
+      "step": 13561
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.171731646780867e-05,
+      "loss": 14.1839,
+      "step": 13562
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.171326921642789e-05,
+      "loss": 16.7245,
+      "step": 13563
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.170922178970666e-05,
+      "loss": 15.9221,
+      "step": 13564
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.170517418767766e-05,
+      "loss": 13.1257,
+      "step": 13565
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.17011264103736e-05,
+      "loss": 17.7654,
+      "step": 13566
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.169707845782711e-05,
+      "loss": 14.8968,
+      "step": 13567
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.169303033007093e-05,
+      "loss": 15.1519,
+      "step": 13568
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.168898202713774e-05,
+      "loss": 14.8909,
+      "step": 13569
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.168493354906024e-05,
+      "loss": 15.5768,
+      "step": 13570
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.16808848958711e-05,
+      "loss": 17.8391,
+      "step": 13571
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.167683606760303e-05,
+      "loss": 16.207,
+      "step": 13572
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.16727870642887e-05,
+      "loss": 14.7845,
+      "step": 13573
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.166873788596083e-05,
+      "loss": 15.4765,
+      "step": 13574
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.166468853265211e-05,
+      "loss": 14.6451,
+      "step": 13575
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.166063900439526e-05,
+      "loss": 16.0122,
+      "step": 13576
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.165658930122295e-05,
+      "loss": 15.2996,
+      "step": 13577
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.16525394231679e-05,
+      "loss": 14.6261,
+      "step": 13578
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.164848937026282e-05,
+      "loss": 16.0376,
+      "step": 13579
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.16444391425404e-05,
+      "loss": 15.4626,
+      "step": 13580
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.164038874003334e-05,
+      "loss": 15.87,
+      "step": 13581
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.163633816277437e-05,
+      "loss": 16.7028,
+      "step": 13582
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.16322874107962e-05,
+      "loss": 16.6752,
+      "step": 13583
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.162823648413151e-05,
+      "loss": 15.1942,
+      "step": 13584
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.162418538281304e-05,
+      "loss": 16.643,
+      "step": 13585
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.16201341068735e-05,
+      "loss": 15.5925,
+      "step": 13586
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.161608265634561e-05,
+      "loss": 15.1632,
+      "step": 13587
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.161203103126206e-05,
+      "loss": 15.9852,
+      "step": 13588
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.16079792316556e-05,
+      "loss": 16.6347,
+      "step": 13589
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.160392725755893e-05,
+      "loss": 14.2631,
+      "step": 13590
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.159987510900478e-05,
+      "loss": 15.3088,
+      "step": 13591
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.159582278602586e-05,
+      "loss": 14.4332,
+      "step": 13592
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.159177028865491e-05,
+      "loss": 16.0953,
+      "step": 13593
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.158771761692464e-05,
+      "loss": 14.7196,
+      "step": 13594
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.158366477086781e-05,
+      "loss": 13.1658,
+      "step": 13595
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.157961175051709e-05,
+      "loss": 14.4027,
+      "step": 13596
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.157555855590527e-05,
+      "loss": 17.575,
+      "step": 13597
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.157150518706503e-05,
+      "loss": 16.3405,
+      "step": 13598
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.156745164402913e-05,
+      "loss": 16.3887,
+      "step": 13599
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.156339792683029e-05,
+      "loss": 17.0192,
+      "step": 13600
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.155934403550126e-05,
+      "loss": 17.1369,
+      "step": 13601
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.155528997007475e-05,
+      "loss": 17.1023,
+      "step": 13602
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.155123573058354e-05,
+      "loss": 14.9002,
+      "step": 13603
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.154718131706034e-05,
+      "loss": 15.3761,
+      "step": 13604
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.154312672953789e-05,
+      "loss": 13.307,
+      "step": 13605
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.153907196804892e-05,
+      "loss": 15.5178,
+      "step": 13606
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.153501703262622e-05,
+      "loss": 14.7575,
+      "step": 13607
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.153096192330249e-05,
+      "loss": 14.5342,
+      "step": 13608
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.15269066401105e-05,
+      "loss": 16.0731,
+      "step": 13609
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.152285118308297e-05,
+      "loss": 16.9562,
+      "step": 13610
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.151879555225268e-05,
+      "loss": 15.5144,
+      "step": 13611
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.151473974765237e-05,
+      "loss": 13.4958,
+      "step": 13612
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.151068376931478e-05,
+      "loss": 15.7458,
+      "step": 13613
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.150662761727269e-05,
+      "loss": 15.5433,
+      "step": 13614
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.150257129155883e-05,
+      "loss": 15.3131,
+      "step": 13615
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.149851479220596e-05,
+      "loss": 17.5727,
+      "step": 13616
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.149445811924685e-05,
+      "loss": 14.6523,
+      "step": 13617
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.149040127271423e-05,
+      "loss": 15.0328,
+      "step": 13618
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.148634425264089e-05,
+      "loss": 16.3856,
+      "step": 13619
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.148228705905958e-05,
+      "loss": 15.2366,
+      "step": 13620
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.147822969200306e-05,
+      "loss": 18.4501,
+      "step": 13621
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.14741721515041e-05,
+      "loss": 15.1007,
+      "step": 13622
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.147011443759547e-05,
+      "loss": 15.6973,
+      "step": 13623
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.146605655030992e-05,
+      "loss": 15.5913,
+      "step": 13624
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.146199848968023e-05,
+      "loss": 15.1815,
+      "step": 13625
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.145794025573918e-05,
+      "loss": 13.5354,
+      "step": 13626
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.145388184851954e-05,
+      "loss": 16.378,
+      "step": 13627
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.144982326805404e-05,
+      "loss": 17.3147,
+      "step": 13628
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.144576451437552e-05,
+      "loss": 16.2504,
+      "step": 13629
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.14417055875167e-05,
+      "loss": 15.5601,
+      "step": 13630
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.143764648751039e-05,
+      "loss": 16.9475,
+      "step": 13631
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.143358721438936e-05,
+      "loss": 14.2406,
+      "step": 13632
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.142952776818638e-05,
+      "loss": 15.9495,
+      "step": 13633
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.142546814893424e-05,
+      "loss": 14.5905,
+      "step": 13634
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.142140835666574e-05,
+      "loss": 15.7227,
+      "step": 13635
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.14173483914136e-05,
+      "loss": 16.0655,
+      "step": 13636
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.141328825321069e-05,
+      "loss": 15.5909,
+      "step": 13637
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.140922794208976e-05,
+      "loss": 15.326,
+      "step": 13638
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.140516745808357e-05,
+      "loss": 14.4842,
+      "step": 13639
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.140110680122496e-05,
+      "loss": 14.4016,
+      "step": 13640
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.139704597154669e-05,
+      "loss": 16.5957,
+      "step": 13641
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.139298496908154e-05,
+      "loss": 16.2018,
+      "step": 13642
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.138892379386234e-05,
+      "loss": 16.8601,
+      "step": 13643
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.138486244592187e-05,
+      "loss": 16.7284,
+      "step": 13644
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.138080092529291e-05,
+      "loss": 15.5743,
+      "step": 13645
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.137673923200828e-05,
+      "loss": 15.8865,
+      "step": 13646
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.137267736610076e-05,
+      "loss": 15.5617,
+      "step": 13647
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.136861532760317e-05,
+      "loss": 15.9093,
+      "step": 13648
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.136455311654831e-05,
+      "loss": 16.6324,
+      "step": 13649
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.136049073296896e-05,
+      "loss": 18.4159,
+      "step": 13650
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.135642817689795e-05,
+      "loss": 15.2947,
+      "step": 13651
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.135236544836808e-05,
+      "loss": 16.5137,
+      "step": 13652
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.134830254741215e-05,
+      "loss": 16.6956,
+      "step": 13653
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.134423947406298e-05,
+      "loss": 16.4067,
+      "step": 13654
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.134017622835337e-05,
+      "loss": 16.8631,
+      "step": 13655
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.133611281031614e-05,
+      "loss": 16.3997,
+      "step": 13656
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.13320492199841e-05,
+      "loss": 15.3754,
+      "step": 13657
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.132798545739006e-05,
+      "loss": 14.9606,
+      "step": 13658
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.132392152256685e-05,
+      "loss": 16.4302,
+      "step": 13659
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.131985741554728e-05,
+      "loss": 14.6275,
+      "step": 13660
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.131579313636414e-05,
+      "loss": 16.5352,
+      "step": 13661
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.131172868505031e-05,
+      "loss": 15.0255,
+      "step": 13662
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.130766406163856e-05,
+      "loss": 13.5611,
+      "step": 13663
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.130359926616172e-05,
+      "loss": 15.462,
+      "step": 13664
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.129953429865264e-05,
+      "loss": 15.6454,
+      "step": 13665
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.129546915914414e-05,
+      "loss": 15.3185,
+      "step": 13666
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.129140384766902e-05,
+      "loss": 16.6134,
+      "step": 13667
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.128733836426014e-05,
+      "loss": 15.0034,
+      "step": 13668
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.12832727089503e-05,
+      "loss": 15.8786,
+      "step": 13669
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.127920688177237e-05,
+      "loss": 14.8528,
+      "step": 13670
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.127514088275914e-05,
+      "loss": 15.6196,
+      "step": 13671
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.127107471194348e-05,
+      "loss": 17.098,
+      "step": 13672
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.12670083693582e-05,
+      "loss": 16.5702,
+      "step": 13673
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.126294185503614e-05,
+      "loss": 16.8939,
+      "step": 13674
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.125887516901015e-05,
+      "loss": 15.0566,
+      "step": 13675
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.125480831131307e-05,
+      "loss": 17.7316,
+      "step": 13676
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.125074128197773e-05,
+      "loss": 17.6995,
+      "step": 13677
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.124667408103698e-05,
+      "loss": 13.7305,
+      "step": 13678
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.124260670852368e-05,
+      "loss": 15.9303,
+      "step": 13679
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.123853916447064e-05,
+      "loss": 15.2182,
+      "step": 13680
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.123447144891072e-05,
+      "loss": 14.589,
+      "step": 13681
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.123040356187675e-05,
+      "loss": 15.4701,
+      "step": 13682
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.122633550340164e-05,
+      "loss": 16.0403,
+      "step": 13683
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.122226727351817e-05,
+      "loss": 15.3016,
+      "step": 13684
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.121819887225925e-05,
+      "loss": 15.1505,
+      "step": 13685
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.121413029965769e-05,
+      "loss": 15.5846,
+      "step": 13686
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.121006155574637e-05,
+      "loss": 16.2644,
+      "step": 13687
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.120599264055812e-05,
+      "loss": 16.2367,
+      "step": 13688
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.120192355412583e-05,
+      "loss": 16.1997,
+      "step": 13689
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.119785429648234e-05,
+      "loss": 15.4649,
+      "step": 13690
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.119378486766052e-05,
+      "loss": 15.0832,
+      "step": 13691
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.118971526769322e-05,
+      "loss": 17.3181,
+      "step": 13692
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.11856454966133e-05,
+      "loss": 14.3757,
+      "step": 13693
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.118157555445365e-05,
+      "loss": 15.2477,
+      "step": 13694
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.117750544124711e-05,
+      "loss": 14.8194,
+      "step": 13695
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.117343515702656e-05,
+      "loss": 15.6573,
+      "step": 13696
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.116936470182487e-05,
+      "loss": 17.0991,
+      "step": 13697
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.116529407567489e-05,
+      "loss": 17.136,
+      "step": 13698
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.116122327860953e-05,
+      "loss": 16.8695,
+      "step": 13699
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.115715231066162e-05,
+      "loss": 16.0787,
+      "step": 13700
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.115308117186406e-05,
+      "loss": 16.2679,
+      "step": 13701
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.114900986224972e-05,
+      "loss": 15.3585,
+      "step": 13702
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.114493838185147e-05,
+      "loss": 16.9632,
+      "step": 13703
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.114086673070221e-05,
+      "loss": 15.9627,
+      "step": 13704
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.113679490883478e-05,
+      "loss": 16.5547,
+      "step": 13705
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.11327229162821e-05,
+      "loss": 17.3728,
+      "step": 13706
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.112865075307704e-05,
+      "loss": 16.7989,
+      "step": 13707
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.112457841925247e-05,
+      "loss": 17.4154,
+      "step": 13708
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.11205059148413e-05,
+      "loss": 15.507,
+      "step": 13709
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.11164332398764e-05,
+      "loss": 14.3934,
+      "step": 13710
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.111236039439065e-05,
+      "loss": 15.5239,
+      "step": 13711
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.110828737841696e-05,
+      "loss": 15.5539,
+      "step": 13712
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.11042141919882e-05,
+      "loss": 16.1078,
+      "step": 13713
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.110014083513729e-05,
+      "loss": 16.345,
+      "step": 13714
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.109606730789709e-05,
+      "loss": 14.4761,
+      "step": 13715
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.109199361030052e-05,
+      "loss": 14.4468,
+      "step": 13716
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.108791974238047e-05,
+      "loss": 15.8254,
+      "step": 13717
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.108384570416984e-05,
+      "loss": 15.5201,
+      "step": 13718
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.107977149570151e-05,
+      "loss": 14.0432,
+      "step": 13719
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.107569711700839e-05,
+      "loss": 15.1313,
+      "step": 13720
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.10716225681234e-05,
+      "loss": 16.5103,
+      "step": 13721
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.106754784907942e-05,
+      "loss": 16.4838,
+      "step": 13722
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.106347295990938e-05,
+      "loss": 16.3945,
+      "step": 13723
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.105939790064615e-05,
+      "loss": 17.3405,
+      "step": 13724
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.105532267132266e-05,
+      "loss": 16.4527,
+      "step": 13725
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.105124727197183e-05,
+      "loss": 15.5696,
+      "step": 13726
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.104717170262654e-05,
+      "loss": 15.4366,
+      "step": 13727
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.104309596331972e-05,
+      "loss": 14.8682,
+      "step": 13728
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.103902005408428e-05,
+      "loss": 15.5793,
+      "step": 13729
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.103494397495313e-05,
+      "loss": 15.3528,
+      "step": 13730
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.103086772595919e-05,
+      "loss": 14.8766,
+      "step": 13731
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.102679130713537e-05,
+      "loss": 13.8002,
+      "step": 13732
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.102271471851462e-05,
+      "loss": 15.0635,
+      "step": 13733
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.10186379601298e-05,
+      "loss": 17.2953,
+      "step": 13734
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.101456103201388e-05,
+      "loss": 15.3068,
+      "step": 13735
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.101048393419977e-05,
+      "loss": 14.7453,
+      "step": 13736
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.100640666672038e-05,
+      "loss": 14.8288,
+      "step": 13737
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.100232922960865e-05,
+      "loss": 14.9502,
+      "step": 13738
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.09982516228975e-05,
+      "loss": 14.0053,
+      "step": 13739
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.099417384661986e-05,
+      "loss": 15.6601,
+      "step": 13740
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.099009590080867e-05,
+      "loss": 14.8513,
+      "step": 13741
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.098601778549682e-05,
+      "loss": 16.9372,
+      "step": 13742
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.09819395007173e-05,
+      "loss": 15.3622,
+      "step": 13743
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.0977861046503e-05,
+      "loss": 15.9747,
+      "step": 13744
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.097378242288686e-05,
+      "loss": 14.8278,
+      "step": 13745
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.096970362990184e-05,
+      "loss": 15.6993,
+      "step": 13746
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.096562466758085e-05,
+      "loss": 16.4523,
+      "step": 13747
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.096154553595685e-05,
+      "loss": 16.9655,
+      "step": 13748
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.095746623506277e-05,
+      "loss": 16.7403,
+      "step": 13749
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.095338676493155e-05,
+      "loss": 16.8001,
+      "step": 13750
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.094930712559613e-05,
+      "loss": 14.9218,
+      "step": 13751
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.094522731708946e-05,
+      "loss": 15.7817,
+      "step": 13752
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.094114733944448e-05,
+      "loss": 17.333,
+      "step": 13753
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.093706719269415e-05,
+      "loss": 17.7207,
+      "step": 13754
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.09329868768714e-05,
+      "loss": 14.8771,
+      "step": 13755
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.092890639200921e-05,
+      "loss": 16.6713,
+      "step": 13756
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.09248257381405e-05,
+      "loss": 15.6261,
+      "step": 13757
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.092074491529823e-05,
+      "loss": 15.8198,
+      "step": 13758
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.091666392351535e-05,
+      "loss": 15.6488,
+      "step": 13759
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.091258276282483e-05,
+      "loss": 17.2923,
+      "step": 13760
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.090850143325963e-05,
+      "loss": 16.664,
+      "step": 13761
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.090441993485268e-05,
+      "loss": 14.5316,
+      "step": 13762
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.090033826763694e-05,
+      "loss": 16.2021,
+      "step": 13763
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.089625643164542e-05,
+      "loss": 15.2323,
+      "step": 13764
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.089217442691102e-05,
+      "loss": 15.6053,
+      "step": 13765
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.088809225346674e-05,
+      "loss": 15.6588,
+      "step": 13766
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.088400991134554e-05,
+      "loss": 16.3576,
+      "step": 13767
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.087992740058037e-05,
+      "loss": 14.872,
+      "step": 13768
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.087584472120421e-05,
+      "loss": 17.0956,
+      "step": 13769
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.087176187325004e-05,
+      "loss": 14.947,
+      "step": 13770
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.086767885675081e-05,
+      "loss": 18.1519,
+      "step": 13771
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.08635956717395e-05,
+      "loss": 16.3809,
+      "step": 13772
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.085951231824908e-05,
+      "loss": 16.4395,
+      "step": 13773
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.085542879631253e-05,
+      "loss": 16.0344,
+      "step": 13774
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.085134510596283e-05,
+      "loss": 15.2712,
+      "step": 13775
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.084726124723293e-05,
+      "loss": 15.3563,
+      "step": 13776
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.084317722015584e-05,
+      "loss": 16.2746,
+      "step": 13777
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.083909302476453e-05,
+      "loss": 16.4532,
+      "step": 13778
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.083500866109197e-05,
+      "loss": 16.5449,
+      "step": 13779
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.083092412917113e-05,
+      "loss": 15.7887,
+      "step": 13780
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.082683942903505e-05,
+      "loss": 17.2441,
+      "step": 13781
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.082275456071665e-05,
+      "loss": 16.1152,
+      "step": 13782
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.081866952424896e-05,
+      "loss": 14.1323,
+      "step": 13783
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.081458431966495e-05,
+      "loss": 16.5781,
+      "step": 13784
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.08104989469976e-05,
+      "loss": 13.5922,
+      "step": 13785
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.080641340627994e-05,
+      "loss": 15.4104,
+      "step": 13786
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.08023276975449e-05,
+      "loss": 15.9097,
+      "step": 13787
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.079824182082553e-05,
+      "loss": 15.1782,
+      "step": 13788
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.079415577615479e-05,
+      "loss": 14.7198,
+      "step": 13789
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.07900695635657e-05,
+      "loss": 16.9669,
+      "step": 13790
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.078598318309124e-05,
+      "loss": 14.9729,
+      "step": 13791
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.07818966347644e-05,
+      "loss": 14.5029,
+      "step": 13792
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.07778099186182e-05,
+      "loss": 15.5138,
+      "step": 13793
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.077372303468565e-05,
+      "loss": 15.671,
+      "step": 13794
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.076963598299972e-05,
+      "loss": 17.0651,
+      "step": 13795
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.076554876359344e-05,
+      "loss": 17.1086,
+      "step": 13796
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.07614613764998e-05,
+      "loss": 15.4216,
+      "step": 13797
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.075737382175182e-05,
+      "loss": 15.7304,
+      "step": 13798
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.07532860993825e-05,
+      "loss": 15.3973,
+      "step": 13799
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.074919820942485e-05,
+      "loss": 17.1673,
+      "step": 13800
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.074511015191187e-05,
+      "loss": 14.3503,
+      "step": 13801
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.074102192687661e-05,
+      "loss": 14.4661,
+      "step": 13802
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.073693353435203e-05,
+      "loss": 14.3566,
+      "step": 13803
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.073284497437121e-05,
+      "loss": 15.2345,
+      "step": 13804
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.072875624696708e-05,
+      "loss": 15.728,
+      "step": 13805
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.072466735217275e-05,
+      "loss": 14.5284,
+      "step": 13806
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.072057829002116e-05,
+      "loss": 17.7487,
+      "step": 13807
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.071648906054539e-05,
+      "loss": 14.8003,
+      "step": 13808
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.071239966377843e-05,
+      "loss": 16.0175,
+      "step": 13809
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.070831009975329e-05,
+      "loss": 15.1187,
+      "step": 13810
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.070422036850305e-05,
+      "loss": 15.3799,
+      "step": 13811
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.070013047006068e-05,
+      "loss": 14.6603,
+      "step": 13812
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.069604040445923e-05,
+      "loss": 16.1661,
+      "step": 13813
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.069195017173172e-05,
+      "loss": 15.242,
+      "step": 13814
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.068785977191119e-05,
+      "loss": 17.2968,
+      "step": 13815
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.068376920503067e-05,
+      "loss": 17.1311,
+      "step": 13816
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.067967847112319e-05,
+      "loss": 14.9718,
+      "step": 13817
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.067558757022176e-05,
+      "loss": 17.5435,
+      "step": 13818
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.067149650235946e-05,
+      "loss": 14.3781,
+      "step": 13819
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.06674052675693e-05,
+      "loss": 16.7682,
+      "step": 13820
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.066331386588434e-05,
+      "loss": 15.2882,
+      "step": 13821
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.065922229733756e-05,
+      "loss": 12.9011,
+      "step": 13822
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.065513056196207e-05,
+      "loss": 16.1443,
+      "step": 13823
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.065103865979088e-05,
+      "loss": 14.6069,
+      "step": 13824
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.064694659085705e-05,
+      "loss": 17.2301,
+      "step": 13825
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.064285435519357e-05,
+      "loss": 14.9987,
+      "step": 13826
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.063876195283356e-05,
+      "loss": 15.3383,
+      "step": 13827
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.063466938381003e-05,
+      "loss": 15.7778,
+      "step": 13828
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.063057664815603e-05,
+      "loss": 15.0586,
+      "step": 13829
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.062648374590462e-05,
+      "loss": 15.8413,
+      "step": 13830
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.062239067708883e-05,
+      "loss": 16.0496,
+      "step": 13831
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.061829744174174e-05,
+      "loss": 14.2294,
+      "step": 13832
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.061420403989637e-05,
+      "loss": 16.1942,
+      "step": 13833
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.061011047158581e-05,
+      "loss": 17.8998,
+      "step": 13834
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.060601673684311e-05,
+      "loss": 16.605,
+      "step": 13835
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.060192283570131e-05,
+      "loss": 14.9633,
+      "step": 13836
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.059782876819348e-05,
+      "loss": 16.2978,
+      "step": 13837
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.059373453435269e-05,
+      "loss": 14.9523,
+      "step": 13838
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.058964013421197e-05,
+      "loss": 15.2186,
+      "step": 13839
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.058554556780444e-05,
+      "loss": 17.3999,
+      "step": 13840
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.05814508351631e-05,
+      "loss": 17.0823,
+      "step": 13841
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.057735593632105e-05,
+      "loss": 13.8313,
+      "step": 13842
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.057326087131136e-05,
+      "loss": 14.5864,
+      "step": 13843
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.056916564016712e-05,
+      "loss": 15.2766,
+      "step": 13844
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.056507024292133e-05,
+      "loss": 16.7234,
+      "step": 13845
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.056097467960713e-05,
+      "loss": 15.4001,
+      "step": 13846
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.055687895025755e-05,
+      "loss": 15.6024,
+      "step": 13847
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.05527830549057e-05,
+      "loss": 14.7457,
+      "step": 13848
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.054868699358462e-05,
+      "loss": 15.2037,
+      "step": 13849
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.054459076632743e-05,
+      "loss": 14.655,
+      "step": 13850
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.054049437316714e-05,
+      "loss": 16.8102,
+      "step": 13851
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.05363978141369e-05,
+      "loss": 15.0596,
+      "step": 13852
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.053230108926976e-05,
+      "loss": 15.7731,
+      "step": 13853
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.05282041985988e-05,
+      "loss": 15.3264,
+      "step": 13854
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.05241071421571e-05,
+      "loss": 16.6571,
+      "step": 13855
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.052000991997776e-05,
+      "loss": 15.2333,
+      "step": 13856
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.051591253209386e-05,
+      "loss": 17.3776,
+      "step": 13857
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.051181497853849e-05,
+      "loss": 15.0816,
+      "step": 13858
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.050771725934473e-05,
+      "loss": 16.5427,
+      "step": 13859
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.050361937454567e-05,
+      "loss": 17.2294,
+      "step": 13860
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.049952132417443e-05,
+      "loss": 16.9291,
+      "step": 13861
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.049542310826405e-05,
+      "loss": 17.0978,
+      "step": 13862
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.049132472684767e-05,
+      "loss": 15.0352,
+      "step": 13863
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.048722617995835e-05,
+      "loss": 15.9912,
+      "step": 13864
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.048312746762923e-05,
+      "loss": 17.0657,
+      "step": 13865
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.047902858989337e-05,
+      "loss": 17.3716,
+      "step": 13866
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.04749295467839e-05,
+      "loss": 14.5302,
+      "step": 13867
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.047083033833388e-05,
+      "loss": 14.6519,
+      "step": 13868
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.046673096457645e-05,
+      "loss": 13.4229,
+      "step": 13869
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.04626314255447e-05,
+      "loss": 16.0565,
+      "step": 13870
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.045853172127174e-05,
+      "loss": 17.7672,
+      "step": 13871
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.045443185179068e-05,
+      "loss": 15.0277,
+      "step": 13872
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.045033181713459e-05,
+      "loss": 14.8181,
+      "step": 13873
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.044623161733663e-05,
+      "loss": 14.9584,
+      "step": 13874
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.044213125242989e-05,
+      "loss": 15.5799,
+      "step": 13875
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.043803072244747e-05,
+      "loss": 17.4863,
+      "step": 13876
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.043393002742249e-05,
+      "loss": 15.3134,
+      "step": 13877
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.042982916738807e-05,
+      "loss": 15.8847,
+      "step": 13878
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.042572814237732e-05,
+      "loss": 15.8861,
+      "step": 13879
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.042162695242338e-05,
+      "loss": 15.615,
+      "step": 13880
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.041752559755932e-05,
+      "loss": 16.5584,
+      "step": 13881
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.04134240778183e-05,
+      "loss": 17.0129,
+      "step": 13882
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.040932239323342e-05,
+      "loss": 16.4146,
+      "step": 13883
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.040522054383782e-05,
+      "loss": 16.6257,
+      "step": 13884
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.04011185296646e-05,
+      "loss": 14.9583,
+      "step": 13885
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.039701635074691e-05,
+      "loss": 15.6039,
+      "step": 13886
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.039291400711785e-05,
+      "loss": 15.4106,
+      "step": 13887
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.038881149881058e-05,
+      "loss": 14.183,
+      "step": 13888
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.038470882585821e-05,
+      "loss": 15.8374,
+      "step": 13889
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.038060598829386e-05,
+      "loss": 14.9001,
+      "step": 13890
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.037650298615067e-05,
+      "loss": 14.9554,
+      "step": 13891
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.037239981946178e-05,
+      "loss": 16.0552,
+      "step": 13892
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.036829648826034e-05,
+      "loss": 15.7055,
+      "step": 13893
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.036419299257944e-05,
+      "loss": 14.2681,
+      "step": 13894
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.036008933245224e-05,
+      "loss": 16.8696,
+      "step": 13895
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.035598550791189e-05,
+      "loss": 15.8234,
+      "step": 13896
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.035188151899153e-05,
+      "loss": 14.7347,
+      "step": 13897
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.034777736572428e-05,
+      "loss": 15.0989,
+      "step": 13898
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.034367304814327e-05,
+      "loss": 14.3366,
+      "step": 13899
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.03395685662817e-05,
+      "loss": 15.856,
+      "step": 13900
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.033546392017267e-05,
+      "loss": 16.675,
+      "step": 13901
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.033135910984932e-05,
+      "loss": 16.1321,
+      "step": 13902
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.032725413534484e-05,
+      "loss": 15.5785,
+      "step": 13903
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.032314899669233e-05,
+      "loss": 15.5895,
+      "step": 13904
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.031904369392499e-05,
+      "loss": 15.5544,
+      "step": 13905
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.031493822707592e-05,
+      "loss": 17.1904,
+      "step": 13906
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.031083259617832e-05,
+      "loss": 13.743,
+      "step": 13907
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.03067268012653e-05,
+      "loss": 15.5958,
+      "step": 13908
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.030262084237005e-05,
+      "loss": 16.4143,
+      "step": 13909
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.02985147195257e-05,
+      "loss": 17.22,
+      "step": 13910
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.029440843276543e-05,
+      "loss": 14.823,
+      "step": 13911
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.029030198212239e-05,
+      "loss": 14.5121,
+      "step": 13912
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.028619536762975e-05,
+      "loss": 15.8414,
+      "step": 13913
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.028208858932066e-05,
+      "loss": 14.6305,
+      "step": 13914
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.02779816472283e-05,
+      "loss": 17.3053,
+      "step": 13915
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.027387454138578e-05,
+      "loss": 16.6677,
+      "step": 13916
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.026976727182634e-05,
+      "loss": 16.854,
+      "step": 13917
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.026565983858311e-05,
+      "loss": 14.7835,
+      "step": 13918
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.026155224168925e-05,
+      "loss": 14.8371,
+      "step": 13919
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.025744448117797e-05,
+      "loss": 14.5712,
+      "step": 13920
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.02533365570824e-05,
+      "loss": 17.7136,
+      "step": 13921
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.024922846943572e-05,
+      "loss": 14.8123,
+      "step": 13922
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.02451202182711e-05,
+      "loss": 14.8712,
+      "step": 13923
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.024101180362175e-05,
+      "loss": 15.5146,
+      "step": 13924
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.023690322552082e-05,
+      "loss": 17.7839,
+      "step": 13925
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.02327944840015e-05,
+      "loss": 14.3147,
+      "step": 13926
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.022868557909693e-05,
+      "loss": 16.0155,
+      "step": 13927
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.022457651084034e-05,
+      "loss": 15.5772,
+      "step": 13928
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.022046727926489e-05,
+      "loss": 14.6979,
+      "step": 13929
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.021635788440377e-05,
+      "loss": 14.9381,
+      "step": 13930
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.021224832629015e-05,
+      "loss": 15.3792,
+      "step": 13931
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.020813860495724e-05,
+      "loss": 13.7513,
+      "step": 13932
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.02040287204382e-05,
+      "loss": 14.6751,
+      "step": 13933
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.019991867276623e-05,
+      "loss": 14.825,
+      "step": 13934
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.019580846197453e-05,
+      "loss": 17.7749,
+      "step": 13935
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.019169808809629e-05,
+      "loss": 14.1375,
+      "step": 13936
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.018758755116467e-05,
+      "loss": 15.535,
+      "step": 13937
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.018347685121291e-05,
+      "loss": 18.6788,
+      "step": 13938
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.017936598827416e-05,
+      "loss": 17.1327,
+      "step": 13939
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.017525496238165e-05,
+      "loss": 13.17,
+      "step": 13940
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.017114377356857e-05,
+      "loss": 15.6493,
+      "step": 13941
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.016703242186812e-05,
+      "loss": 17.1035,
+      "step": 13942
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.016292090731348e-05,
+      "loss": 14.8164,
+      "step": 13943
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.015880922993789e-05,
+      "loss": 15.9849,
+      "step": 13944
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.015469738977452e-05,
+      "loss": 16.6502,
+      "step": 13945
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.015058538685658e-05,
+      "loss": 13.9393,
+      "step": 13946
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.014647322121727e-05,
+      "loss": 15.95,
+      "step": 13947
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.014236089288984e-05,
+      "loss": 16.4666,
+      "step": 13948
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.013824840190745e-05,
+      "loss": 17.1791,
+      "step": 13949
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.013413574830331e-05,
+      "loss": 15.8674,
+      "step": 13950
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.013002293211066e-05,
+      "loss": 14.4145,
+      "step": 13951
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.01259099533627e-05,
+      "loss": 16.6763,
+      "step": 13952
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.012179681209263e-05,
+      "loss": 14.3393,
+      "step": 13953
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.011768350833369e-05,
+      "loss": 14.4631,
+      "step": 13954
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.011357004211907e-05,
+      "loss": 14.8855,
+      "step": 13955
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.0109456413482e-05,
+      "loss": 15.984,
+      "step": 13956
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.010534262245571e-05,
+      "loss": 15.3962,
+      "step": 13957
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.010122866907341e-05,
+      "loss": 15.4002,
+      "step": 13958
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.00971145533683e-05,
+      "loss": 15.9867,
+      "step": 13959
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.009300027537363e-05,
+      "loss": 14.6302,
+      "step": 13960
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.008888583512262e-05,
+      "loss": 16.0202,
+      "step": 13961
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.008477123264848e-05,
+      "loss": 16.8194,
+      "step": 13962
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.008065646798446e-05,
+      "loss": 15.0656,
+      "step": 13963
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.007654154116377e-05,
+      "loss": 15.0036,
+      "step": 13964
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.007242645221964e-05,
+      "loss": 14.3688,
+      "step": 13965
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.00683112011853e-05,
+      "loss": 16.1699,
+      "step": 13966
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.006419578809399e-05,
+      "loss": 15.88,
+      "step": 13967
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.006008021297895e-05,
+      "loss": 14.7832,
+      "step": 13968
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.00559644758734e-05,
+      "loss": 15.1673,
+      "step": 13969
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.005184857681057e-05,
+      "loss": 14.007,
+      "step": 13970
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.00477325158237e-05,
+      "loss": 15.5728,
+      "step": 13971
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.004361629294607e-05,
+      "loss": 16.7911,
+      "step": 13972
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.003949990821085e-05,
+      "loss": 15.1815,
+      "step": 13973
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.003538336165133e-05,
+      "loss": 15.0035,
+      "step": 13974
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.003126665330073e-05,
+      "loss": 15.3824,
+      "step": 13975
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.002714978319232e-05,
+      "loss": 15.8933,
+      "step": 13976
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.00230327513593e-05,
+      "loss": 15.3997,
+      "step": 13977
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.001891555783497e-05,
+      "loss": 15.1632,
+      "step": 13978
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.001479820265252e-05,
+      "loss": 15.866,
+      "step": 13979
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.001068068584524e-05,
+      "loss": 15.2775,
+      "step": 13980
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.000656300744636e-05,
+      "loss": 16.4095,
+      "step": 13981
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.000244516748916e-05,
+      "loss": 16.284,
+      "step": 13982
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 6.999832716600685e-05,
+      "loss": 14.7197,
+      "step": 13983
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 6.999420900303273e-05,
+      "loss": 14.9449,
+      "step": 13984
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 6.999009067860002e-05,
+      "loss": 14.5861,
+      "step": 13985
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 6.998597219274199e-05,
+      "loss": 15.1495,
+      "step": 13986
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 6.998185354549189e-05,
+      "loss": 16.2117,
+      "step": 13987
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 6.997773473688299e-05,
+      "loss": 13.5766,
+      "step": 13988
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 6.997361576694854e-05,
+      "loss": 16.0274,
+      "step": 13989
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 6.996949663572181e-05,
+      "loss": 16.5608,
+      "step": 13990
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 6.996537734323606e-05,
+      "loss": 15.0861,
+      "step": 13991
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 6.996125788952456e-05,
+      "loss": 16.4329,
+      "step": 13992
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 6.995713827462056e-05,
+      "loss": 14.7086,
+      "step": 13993
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 6.995301849855734e-05,
+      "loss": 15.4827,
+      "step": 13994
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 6.994889856136816e-05,
+      "loss": 15.9136,
+      "step": 13995
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 6.994477846308632e-05,
+      "loss": 14.3597,
+      "step": 13996
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 6.994065820374505e-05,
+      "loss": 14.0986,
+      "step": 13997
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 6.993653778337763e-05,
+      "loss": 15.2296,
+      "step": 13998
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 6.993241720201737e-05,
+      "loss": 17.5899,
+      "step": 13999
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 6.99282964596975e-05,
+      "loss": 15.7668,
+      "step": 14000
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 36042,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 2000,
+  "total_flos": 1.6510394588092747e+20,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}